👉│feedback-general

[runic zephyr]

That's very painful, today i'm facing this. I thought only me. But really don't want any other option than deep research. Im get irritated to manually change it for every query

[blissful jungle]

Perplexity's official team is really crazy, always doing these pointless upgrades. The default to Sonar is definitely just to promote their own model. And GPT-4.5 is gone for Pro users too, what the hell are they doing?! @covert sun

What's the reason for creating an operation that unnecessarily raises the cost of user operations?

[sturdy geyser]

So not hating on Perplexity but seems like GPT-4.5 model isn't being used even selected in the settings... and your 10 messages quota decreases for GPT-4.5 even if it's Perplexity Pro messaging back... Is it really ethical?

Crap didn't see above messages but hey one more proof

I remember year to 2 years ago Perplexity was one of those underdogs coming after top dogs.

[fiery shard]

• Feature Reuest *
  Can we please add mcp to perplexity windows app like claude app?

[sturdy geyser]

Guys it'd only perplexity pretending to be other models... claude test

I canceled my subscription

[reef lodge]

Feedback on Deep Research: it’d be great to have some indication that the prompt is still running in the background. On iOS, it’ll often show a network connectivity issue and in the library there’s no indication as to what is going on with the prompt, plus are turned off for personal reasons. It’d be good to have an in-app signal on unread notification count on the app. Besides network connectivity, the progress summary freezes then it’s not clear what’s going on since there’s no error message.

Otherwise this has been a wonderful product with valuable day to day usage.

[tropic raft]

That's not intended behavior. Mind DM'ing me a video of this?

[tropic raft]

?

[tropic raft]

thanks let me find out what's going on

[tropic raft]

@loud agate @mighty fox appreciate you flagging. We made a change yesterday that follow-ups will be changed to to "Auto" by default.

Our reasoning is that we've seen most follow-ups don't require Deep Research or Reasoning and is actually just making you wait longer when most of the time, you're looking for a quick follow-up response. Designed to make follow-ups a lot faster and simpler.

[tropic raft]

What type of searches are you doing that you're always on Deep Research? no judgement just genuinely curious what you're looking for that you're willing to wait that much time for each answer

[runic zephyr]

It's helping my studies and my PHD. I need to go through the number of materials and the number of facts and make a decision out of it. Can't answer more deeper to your question as it's my personal stuff. But hope you got my reason for deep research usage.

[loud agate]

This change is really inconvenient. Not only are Reasoning and Deep Search affected, but even Pro keeps getting reset to Auto. And as a user, I have the ability to choose the language model that suits me. This change just forces me to manually adjust it every time, which is really annoying.

[red steeple]

I don't see GPT-4.5 either

[runic zephyr]

And perplexity ranks second in humanity last exam benchmark

And perplexity is having unbeatable benchmark in Q&A evaluation when compared with any other AI provider.

So i have lot of confidence on perplexity and its research which is provided for 20USD, in a human possible budget

[austere hedge]

for me Perplexity seems to never do web searches for follow-ups now, even if I manually select 'pro' again, is it intended?

[blissful jungle]

I don't have this

[blissful jungle]

To be honest, I don't agree with this change. First of all, I don't understand how you conducted the research to arrive at this conclusion. Even if most people choose 'Auto' instead of 'Reasoning', you shouldn't deprive the minority of their preferred usage and force us to use 'Auto' mode by default.

We naturally have our reasons for preferring 'Deep Research' or 'Reasoning', such as when we need to do coding or research work. Those who are accustomed to using 'Reasoning' will naturally continue to use it, and those who primarily use 'Auto' are unlikely to switch to ' Reasoning'.
This is completely unnecessary official intervention, it's totally redundant

[warm trout]

I have an issue. This afternoon I gave Perplexity’s AI bot an http link to an online PDF and asked it to extract all the text from there. It did the job just fine. This evening I did not have that text handy, so I asked the bot in another thread to do it again. The response I got was that Perplexity cannot read online files dynamically, claiming that earlier it must have obtained the text from somewhere else on the web. That text has never been anywhere else on the web! But the bot insisted it cannot read specified online files dynamically. I pointed out that, as I was deciding whether to use Perplexity, I asked many people whether Perplexity can do that — and everyone said yes. I’m sure I’ve seen advertising that said it could do that, but it would take a search to find it again. Can Perplexity read a specified file online as everyone says it can, or can’t it? Is there false advertising?

[wooden parcel]

So I discovered tonight that Deep Research reports are a great length to listen to while working out. I know, power user bro right here. Anyhow it doesn't work out because the audio on Android always dies, like most of the way through! You did such a great job fixing the hallucinations on DR I bet you can knock this one out too real quick. Thanks for your hard work!

[shut arch]

Hey @wooden parcel!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[languid stratus]

Hello, upon my renewal, I did not get the $10 discount on one successful (paid) referral

[mighty fox]

I would assume it is more about saving tokens than about UX. Let's figure out a better way to save tokens? How about limiting the number of reasoning answers every 6 hours or so? This way you won't get hit by a huge API bill from Anthropic/OAI and your users still can use reasoning without having to click on it everytime. Win-win?

[noble monolith]

please contact [e-mail]

[charred patio]

I agree that the default is automatic, but I would like to be able to customize it.
I'm working on a project that involves a lot of thinking, such as researching wide and deep data and reasoning about psychology, and it's inconvenient to have to keep choosing.

I think people who use pro do so for a reason.

[gray hare]

I think this reason not really make sense, if the user wants fast-shorter-simple answers he/she can always change it(before updated: just one time), but if the user wants depth-longer-detailed answers he/she has to change it(after updated: every time). It's easy(even without AGI) to make a conclusion that: The benefit of setting the default to "AUTO" is not really for the users, but more for the company. Finally, why not just give users the option to decide what the default is?

[unique nexus]

Guys, why is the "high" deep research not available on desktop? Sometimes it feels like both web and desktop versions are very different and away from each other.

[quartz pivot]

The voice input feature on the windows app doesn't work with german (and i guess other languages), only with english. If i say german words they get translated to english, but i would prefer to see my german words in the input fields.

[noble monolith]

known bug, thank you for reporting

[sonic jacinth]

Look at a browser add-on named Complexity. It customized the the interface to make it more friendly. If It doesn't already it soon will have an option to make deep search static.

[next knoll]

Agreed. Upvote and bump the Pro feedback request here https://discord.com/channels/1047197230748151888/1352672220949778553

[heady plover]

Chat gpt ran a maze on me, hurt me, got months of logs, proof, need legal help anonymously.

[dense zenith]

Why not give the user the option of showing potential related questions at the bottom of a search query, instead of requiring five per input

[fading storm]

• forcing Pro users to use Auto, despite paying for Pro
• forcing free users on Auto to use their daily Pro allowance

what even is perplexity's game plan here? to drive away everyone using the site?

[tight flax]

Dunno. But LLM always has a wow factor for about 1 weeks then it's nerfed. I spent several hundred with this company already.

[serene tulip]

Hi, can you create channels for enterprise customers, especially for asking/help on enterprise related features (search on your own data etc)?

[pliant wyvern]

Why pplx removed the high mode of deep research?

[left sleet]

I am an enterprise customer, but I do not have access to this channel. How do I gain the access?

[mighty fox]

I can't access that channel, how can I get invited to it? Is it a "pro" channel or something?

[mighty fox]

...aaand Claude just announced web search, so another reason NOT to use Perplexity anymore

[compact harness]

Suggest a short name for Perplexity eg px - easy to call again and again easily

[runic zephyr]

Plex – Clean, cool, and intuitive (like “plex of info”).

PX – Your original suggestion, very snappy and techy.

Perpy – Playful and friendly, if you want a less formal tone.

Plx – Edgy and even shorter than “Plex”.

Perx – A hybrid of “Perplexity” and “PX”, sounds futuristic.

PZ – If you want to go ultra-short, similar to “GPT” or “AI”.

PeeX – Phonetic version of PX, feels more casual.

[noble monolith]

https://discord.com/channels/1047197230748151888/1111765634267742408

[wooden parcel]

I love the new DR, but I want to encourage you to make the output length like 50% more, even if you have to limit use somewhat. I can see everytime I use it, it collects useful information that it just can't fit into the output window. Limiting such a powerful search tool like this feels wasteful! If I'm willing to wait 10-20 minutes for the report, I won't mind reading 12 pages.

(I'm talking about the High DR. The standard I think is a fine length!)

[runic zephyr]

i cant see new DR, i see the same one as before

[deep ermine]

anyone else having problems right now?

I'm not getting real-time replies. Need to keep refreshing and eventually the reply shows up

[golden geyser]

iphone app crashes on a space thread, that is most likely consuming too much memory

[somber flower]

This is horrendously inconvenient. I agree with someone else who already said people who are going to use Auto are unlikely to switch from it.

You cite time saving reasons, but I am wasting far more time having to constantly switch back to the model I want to use.

Your idea of faster and simpler is slowing down and complicating my workflow.

It's making me regret my annual subscription. If this is going to be my user experience, I absolutely won't renew.

[compact harness]

Yeah, Plex is good. Lets propagate it as short name for Perplexity

[grand cobalt]

Peex is tuff

Also I just bought the Claude subscription for the new web search feature and it sucks

[grand cobalt]

Joke btw

[runic zephyr]

Yeah

[fading storm]

i keep coming back to perplexity out of habit, but i think i'm now at the point where i won't bother anymore

you're telling me
perplexity DECIDED to use up one of my pro searches
because i needed to quickly whack on https://chromewebstore.google.com/detail/ on the start of each string

WHY

ok yeah it did the job, but it did not have to PRO SEARCH???

[orchid lake]

The new UI. When I ask a follow-up it turns into auto mode automatically, so I have to switch it to my mode of choice (Mostly use reasoning mode). Sometimes there is some annoying things happen from this UI, like I just wanted to use reasoning for follow-up question, but then I just hit enter before I had changed the option, so I have had to resent the question again.

[lyric nimbus]

yeah, probably much smaller llm could do the job. seems even those free ones on https://labs.perplexity.ai could work. this is from the r1 finetune (sonnet didnt do best job with extracting those ids from your screenshot...)

[grand cobalt]

Tf is your banner weirdo

[fading storm]

my problem is more of, perplexity unsetting pro searches on those who pay for it, and setting pro searches for those who don't pay and want to use their free quota wisely

it's just such a bizarre take, not sure why the vision is but it's driving me nuts

[grand cobalt]

Bro just use andi until they fix it (perplexity is better I’m just saying Andy until they fix the problems they have) admins please don’t revoke my subscription please 🙏

[somber flower]

Having to resend because I forget that I have to switch back to Reasoning EVERY TIME is super annoying and happens a lot.

[wintry orchid]

ooh so I am not the only one deeply disappointed by last month features released and changes to the service? continuing like this i think i will not renew after 2 years of usage

[heady plover]

lol perplecity sucks, alarmingly frequent false and contradictory responses and constantly telling you you don't understand something only to just re-word your response proving you were right to begin with, Those that program this AI model need to work on it big time, it is one of the most flawed LM models i've ever used.

[lyric nimbus]

well, I was using pro with sonnet for almost everything. now when it shadily switches on its own to "Auto", I just select reasoning sonnet or o3-mini out of spite, to waste pplx money 😅

[sacred compass]

First off, I LOVE Perplexity! Thank you for a great product!

I'm a HEAVY web app user, and the recent change where the "Auto" model is set for every chat is SUPER annoying.

I understand this was likely done to mitigate costs, as the Auto model uses Perplexity's Sonar model more often than anything.

I have to change the model for every individual message, which is very frustrating.

Can we please find a common ground for this?

I'd love to go back to the default model being set in the settings...

But if you're trying to minimize costs... could you keep the thread the same model AT LEAST please?

[sacred compass]

Good to see someone else shares this frustration too!

[molten mountain]

does anyone know how to remove these Categories ?

These annoying news always appear on my home page.

I don't know how to delete.

[lyric nimbus]

the "Auto" model is set for every chat is SUPER annoying
strangely, I also thought it sets to Auto after every query I sent. but I seem to have occasionally hit a weird streak when it remembers my previous mode correctly
I understand this was likely done to mitigate costs, as the Auto model uses Perplexity's Sonar model more often than anything.
I think it used to give me always sonar, then I think it was always sonnet (when the model preference was in settings). now it most of the time picks sonar, but I remember forgetting it on Auto with a more complex (longer) question and the response felt more like sonnet or 4o than "small" sonar (llama 3.3 70B?)

[lyric nimbus]

I am a bit confused. isnt that just "x" button on a specific item?

if you mean hide from the home page, that wasnt possible (I use perplexity helper to hide it, I think complexity can do that too)

[molten mountain]

It's not the Watchlist inside, but the entire Sport Category. Even if the Sport Category Watchlist is empty, it still pushes sports news to my homepage.

[lyric nimbus]

yes, I think when it is empty in settings, it means for perplexity to show everyhing or maybe it is target based on what perplexity knows about the user? last I checked it wasn't possible to disable it (only hide via userscripts or browser extensions)

personally I would prefer more customization on perplexity, or even some advanced mode, to be able to set more technical stuff, like temprature and other parameters for llms, or custom/none system prompt in some writing/creative mode. I think uncovr had this handled better - a bunch of toggles in settings to show/hide calendar, news and other widgets on homepage

[molten mountain]

Still thanks for your answer.

[lofty jewel]

Why does every message have suggested prompts

[willow plank]

complexity

[patent rampart]

I set it to not search the web, Sonnet. It then proceeds to search the web (confidential info btw) and switches to sonar (even though sonnet is selected).

[pastel zenith]

except tylis on voice mode, none of the others are working on mac

[willow plank]

Did perplexity just die

[wicked swallow]

Yes

[wild dagger]

Is everybody having issues with perplexity currently?

Keeps telling me something went wrong try again

[heady plover]

This is the message I get when using Perplexity features.

[willow plank]

derplexity

[ruby canyon]

Yeah... Having issues with spaces and threads not being available... Lost a number of threads as well.

[heady plover]

Spaces seems to not appear at all, and nothing is responsive with this message above in brown background.

[ruby canyon]

Also, I can't seem to figure out how to make a page from a space with lots of threads. I can make a page from a single thread, but not a space.

[rocky estuary]

Same thing here

Status page is not showing that.
https://status.perplexity.com

[heady plover]

Its just Perplexity, here is the result from other websites that check the status of websites.

Here is another one.

[rocky estuary]

Yes, it's down and can see that. I sent the status page link to show is not updated.

[heady plover]

Good thing, so people don't get misled by Perplexity status from the official website.

[silk spire]

please revert the change that it changes to auto mode after every query, its super annoying and inconvenient. there is a reason im using this specific mode.

[sly herald]

Wen deepseek v3 0324

[blissful jungle]

@tropic raft @covert sun Why haven't you rolled back the update regarding the "default automatic mode" yet? Many users gave feedback last week. What exactly are you officials doing?

This is truly truly annoying

[hybrid halo]

Please modify the prompt output on Discover articles to not use the phrase “marks a significant” so often, or at all. It’s on just about every article and it’s driving me mad.

[heady hornet]

Voice mode on ios keeps repeating i tell it and ask it will you stop repeating it goes sure will and it does it again literally 5 times IN A ROW

[wispy vale]

I just want reasoning mode but always creating new thread or prompt uses auto. This is frustrating a bit.

[bleak helm]

Three most annoying app issues in Perplexity- please allow us using web app on iOS experience because the iOS/Mac apps are so very much limited (web apps work on Mac) :

1. The apps really give a bad user experience compared to the web frontend. They are way more buggy (esp. MacOS) and they lack changing the LLM model on the fly .(independently) for "Pro/Reasoning/Deep Reasoning" which is so much worse that I wonder to stop using Perplexity on iPad at all! Theoretically one can install Perplexity as a web app (but badly supported on iOS because of silly entry page), so there is no reason for this problem at all !!! Stop making apps when they are so limited ! Change to web apps
2. As said: Using web app on iOS could be an easy way out: At least using a web app on iPad should be promoted as best option for power users. But you modified the entry page so that using the web app (installed e.g. on chrome) is not convenient because page loses context after switching tasks
3. In the web I can see a list of all sources by clicking on the rightmost source icon. This could be an extra menu entry (bad UI). But this does not work in apps too which is annyoing.

[patent rampart]

I'm literally gonna scream - since last week it keeps searching, even when search is switched to off. I'm using unpublished research data, I can't have the documents contents used as search queries. It's 100% due to this default automatic BS.

This is for follow up queries btw

[sinful depot]

when using o3 mini, math and certain notations dont come out nicely - subscripts are replaced with '_'

[sudden belfry]

Site and MacOS app erroring out for me

Keeps saying "try again"

[digital willow]

guys can u help me, last night my threads were working fine, but today it cant be accessed and went missing in my space

[shut arch]

Hey @digital willow!

If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.

[languid kindle]

#💻︱feedback-desktop Hey guys! First of all... Love your product! It's really great. However, there is one major downside for me, that makes your mac os app very annoying. Please, PLEASE, make it work in background. I don't want to see the app icon all the time, let me just bring the text window just using the shortcut and place the app icon in the background. It will be soooooo much better!
Once again, love your app, hope my feedback will be useful!

[heady plover]

For now, you can know that the underscore is acting to indicate subscript.

The picture will show you what does this mean in normal math function:

• d_t + i_t = af(k_t)

Hope this help as a workaround.

[visual bough]

Hey @rugged plover!
Please use #😎│cool-projects
Thanks 🙏

[whole compass]

Could you add a subscription payment in euro directly on your website? So that in Europe, I don't have to pay in USD (avoid the Exchange Fee).

[river birch]

Using the desktop app on windows for some days now, I have one big issue: You can't switch between different windows while editing a new prompt - if you do so, your text is lost and you have to start from the beginning. This sould be fixed in next versions...

[sweet laurel]

Hello! Does someone knows if the app for ios its with some problem? Its not working for me. Nothing.
I checked and theres no update, at least it doesn’t appear any update to me.
I saw that the web app was having problems but i cant find anything about the app.
The last time this happened it took 3 days to come back to normal. Which is kind absurd, I never saw an app spend so much time not working like this. Specially a paid one

Does someone knows the newest version of ios? Just to be sure. But it really doesn’t appear any updates on apple store

[vital cape]

will math and video mode be added back in web?

[graceful phoenix]

#🌐│feedback-web Hey, i just joined to share just this feedback. In the latest update i find, that after selecting the thinking or reasoning or deep search the model gets switched back to the auto. This is extremly irritating because as i found out the auto model usually takes the (probably cheaper) "stupider" model. It then kinda ruins the whole session because evan after i select the better details, sometimes i have to start from scatch. I love the product but this user experience is kinda ruining for me. What do you guys think?

[cursive bluff]

I realise you guys has make the deepsearch much more depth and longer in term of their process, i enjoying the result so far.
Keep it improve team.

[sharp shell]

Do you guys have any affiliate program or creator partnership programs? If you do, can you put me in contact With someone. thank you

[halcyon osprey]

Defaulting away from websearch when websearch is selected nearly destroyed perplexity, real sad

Was nice while it lasted

But they taught others how to do it right.... someone will pick up the torch one day

[devout nest]

Can you bring in selection of reasoning models for rewrite option in existing thread like before.

[simple marten]

Please FIX the horrendous image generation UI. Thank you.

[plucky moon]

Can we have more voices for the voice chat mode. At the moment there all American which is fine but I would like some British ones and some from other regions.

[charred aspen]

I think they turned off image gen or modifying attachment image. Can’t find any options on iOS or web. @toxic forum isn’t it? Midjourney as an option is also gone

[toxic forum]

You're correct that there have been some changes to image generation features recently. The web app for Midjourney has experienced issues, particularly on iOS devices. As for ChatGPT's image generation, it's still available but may have limited functionality on mobile devices. The removal of Midjourney as an option suggests ongoing updates or changes to the available AI image generation tools.

If you're having trouble accessing these features, you might try using a desktop browser or checking for any app updates. It's also worth noting that AI image generation tools are rapidly evolving, so temporary disruptions or changes in availability are not uncommon.

[charred maple]

I use this tool daily. 5 star review. However, there is a crystal clear political bias and agenda in the responses that are generated from perplexity. What are developers doing to fix this problem? An echo chamber is NOT good for anyone.

[gentle verge]

I would love to be able to connect to MCP servers in Enterprise Pro! Especially at the space level.

[pure plover]

Can Perplexity for Mac remember its window position when reopening, please

[silent flame]

Wtf is window position

[wintry orchid]

Funny that even the status site is telling is not down lol 😂 https://status.perplexity.com/

[deep chasm]

Good morning guys, I can't select the template to use, I had also preset to use Claude 3.7, but it prompts me in the prompt to use the selection from there.
Auto NEVER gives me the answers I'm looking for and what's more, AUTO mode always gives me wrong solutions.
What's more I am also forced to use auto mode, any other mode I choose always returns me to auto.
Please put it back the way it was before.

[tired nimbus]

I am having the same issue, there seems to be a disconnect what revolut is offering and how perplexity is implementing it, for the time being I just created a new account but this is not ideal, in one year I would have to do the same

[vague rapids]

I agree there seems to be a disconnect. Other subscription based services offered through Revolut do not have this issue and honour the benefit as they should. This issue is specific to Perplexity. I still have not heard anything back here in Discord but thankfully email support has added the code to my existing account. They said it will work for 2 years, god bless.

[deep chasm]

OK, the problem is solved in part, but I still can't force Claude as the default AI.
I don't want the machine to decide how I should use the prompt, I want it to be me.

[thin willow]

Hello, my Perplexity Pro is not working. It erased all my spaces and libraries and when I go to create a new thread, it does not allow me to use the pro features of changing the search engine from Auto to Pro or Reasoning

[rare nimbus]

Outage

[thin willow]

Thank you

[rare nimbus]

np

[hasty elbow]

Hello good morning. Apparently is a problem perplexity is having. I also have Perplexity Pro and cant see my libraries. Any solution?

[covert sun]

Hey @hasty elbow! Please take a look at: #🔔│status message
Sorry for the inconvenience, it will load, if it hasn't, yet.

[hasty elbow]

Thank you very much sir, Perplexity is so awesome that i can barely function without it. You guys are doing a great job. I will wait.

[drifting pulsar]

Perplexity team, guys, I've noticed that the auto model selection has led to subpar responses, I have to switch to proper models each time, which has been very frustrating.

Additionally, despite being on the Pro plan, I encountered a message stating, "0 enhanced queries available today," while I didn't use it that much today, which was unexpected. Could you please provide some clarity on these issues?

[crude frigate]

I´m a Pro User for the last 6 months. Today I opened Perplexity and I´ve lost all my information. there is no library no spaces... NOTHING!!
What can I do??

[keen anchor]

Perflexity, please please fix your model selection. I have tried everything, Firefox, Chrome etc. It keeps dispersing. Whoever opts for the idea to remove this option in the settings menu must get a deduction in salary!!!!

[river apex]

i cant use any models except auto

[timber atlas]

Also, I can't use perplexity and the story is gone

[nova portal]

is everything in perplexity broken? all my spaces and history is missing

looks like mutiple pleople having this issue. please add your comments here #1355225126899286198 message

[noble monolith]

#🔔│status message

[vapid kraken]

Perplexity canceling subscription then ignoring support mails is wild. Not very satisfied with support

[vagrant isle]

I think they have some big issue and no one have time to explain us what is going on xd

THIS IS DISRECPECFULL XDDDDD

[wicked swallow]

Shots fired

[glad pilot]

would it be possible to get a specific tab to be able to generate images in. i don't do much of it, but i find it quite frustrating to have to do a search or "chat" and then click to generate, and invariably the llm has responded that it can't generate and waffles on about how you'd go about generating or drawing the image, and that gets put into the prompt (though that does seem to have improved somewhat), so i find myself editting the promt too. -- its nice to have multiple generator models, but just a bit frustrating to actually use them like it is atm. would also be nice to be able to switch the model on the fly rather than going into settings

[rigid hare]

Perplexity: The reason I can't read the attached Excel file is technical. As a text assistant, I cannot open or analyze files directly from external sources, including Excel documents.
Okay, this is disappointing. I use Perplexity Pro and lost more than 2 hours trying to make it analyze attached excel sheets (one at a time), cutting them in smaller and smaller in size and number of rows. And after all the efforts the answer I get asking why the task cannot be completed is this.

[barren brook]

I just tested an Excel (.xlsx) file with multiple sheets, including a pivot table, and Perplexity had no trouble reading it. (Pro - Claude 3.7 Sonnet)
Does your file have the .xlsx extension, or is it .xls, an older version that Perplexity does not support?

(I also tried in a Space with the file attached as a source, and it works as well—Perplexity was able to read it without any issues.)

[rigid hare]

it is .xlxs and the worst is that the model, automatically picked by Perplexity, is pretending to do an analysis and provides some relevant comments, like 10%, but the rest is imagined (hallucinated) and when asked to explain its inferences, excuses and explains that it cannot in fact read the file attached
what's more, when asked to read the text in specific cell, it did and cited it, but exposing the discrepancy between what citations it extracted in the previous step and how it analyzed the contend answering to specific question and the full text it in that cell (that it found when asked), it claimes that it is not able to read the file attached

[barren brook]

"automatically selected by Perplexity"
Do you have a Pro subscription or not? If so, choose Pro - Claude 3.7 Sonnet.

[rigid hare]

yes, I do, I am a Pro subscriber for almost an year now

[rigid hare]

is your recommendation to use Claude 3.7 for such tasks?

also, the language is not English

[barren brook]

Here, in Perplexity, yes.

[rigid hare]

thanks, I have been using it preferably, but still experimenting and comparing with the other options for different tasks and over relied to the Auto function to pick the most suitable one for specific exercises

[barren brook]

I never use the auto mode. If you're familiar with the models, you're fully capable of making your own choices. Auto mode can be helpful for less experienced users, and it's a good initiative from Perplexity.

[rigid hare]

btw, am I wrong or until recently it was possible to see the model used for every prompt while now I cannot easily find where to look for that (no matter if I explicitly picked the model or left it to Perplexity to auto-choose, which in my case is rare)

[barren brook]

Yes, in Perplexity, you need to hover over the button at the bottom right of your response to see the model used.

Personally, I use the "Complexity" extension, which provides more control over various aspects of the Perplexity interface. Here, for example, is the model selection menu for CPLX and PPLX. (https://www.cplx.app/)

[rigid hare]

Thanks, only if manually set (e.g. to Claude 3.7), not if in Auto mode

[rigid hare]

thanks, I'll check it

[vivid aspen]

While removing the response_format parameter from Tier 0 recently, did Perplexity notify its API users? If not please notify in-advance of any format changes to tiers.

[nova plaza]

Hey everyone, I'm new here and I have one question. Would it be possible to have a similar feature to the chat gpt client to use the voice to write down a message ?
Not a discussion with the AI, not a thing that pops up and you have to hold to speak.
Something that I start, and end when I'm done describing my ideas/questions etc.
It's convinient when walking and thinking.

[safe frost]

When you leave the window in the center of the screen, it should open in the center of the screen

[silent flame]

Why tho

[chilly jacinth]

ABSOLUTELY, I remember back when i used chatgpt it was such a good feature, saved me so much time and was such a time and effort saver

[silver hazel]

I don't mind all the changes, I am not a power user and use perplexity as an improved search engine. I like the new update that's merged Pro and Reasoning. What I would like is that the android app matches more quickly the changes of the web version. It is like 2 different products

[obtuse dawn]

I noticed that the text "추론" (translated as "reasoning" in English) in the model selection menu is displayed vertically. However, for better aesthetics and consistency, it would be more visually appealing if the text were displayed horizontally instead of vertically. This adjustment would align better with the overall design and improve readability.

[spiral sapphire]

New selector is gorgeous. Thank you

[wooden parcel]

Is there a way in the new Pro "Best" model selector to choose the reasoning model that is used when a reasoning model is needed? I like not having to choose the model each time, but I would have more confidence in using that option if I knew it didn't use R1, for example, because I know from experience this is a very unreliable model. As it is, I can't even tell after the fact which model was used.

[tropic raft]

This was never available for Tier 0,1,2. We simply made the error message more clear in the API call and website

[vivid aspen]

Thanks. Looks like this was an error on the replit co-pilot side. It had included response_format parameter without recognizing my tier. The error showed up first time when I was doing a client demo yesterday…

[tropic raft]

We are making a change to this soon though. Stay tuned

[hoary gorge]

Im not a developer or anything, but while you wait for an official answer you could try to use the text to speech option from your phone. Most of them work like your request!

[nova plaza]

Mine is bringing up Google, and besides the poor quality of the transcription, when I try to take a breath, it automatically sends the unfinished phrase for search. So it's relatively unusable.

[blazing epoch]

Does the keyboard on your phone not have the toggling feature, where you can take your breaths in between?

Mine has it that I can select it, say a few things and if I pause, it will turn off, but it won't send the query.

[nova plaza]

Look, basically, I am a polyglot and use many languages. Google and Samsung's transcriptions do not work well, especially when switching between languages like English, French, Russian, and Armenian. They cannot handle this... the quality of transcription in English is good for most of these tools, but it is bad in other languages.

[wicked swallow]

Old news

[nova plaza]

I just tested them, the quality of whisper is sadly better from what I have tested myself.

[brave anvil]

Can the upper limit of a space's storage capacity be expanded, for example to 100 files (or can multiple spaces be linked for in-depth research)

[normal falcon]

Wtf is deep research high

[wicked swallow]

It's Google search after joint

[barren brook]

D.R.H. was active for a few days before being removed. The CPLX extension had implemented it directly. Since it is no longer available (for now) and there is a new UI on PPLX, the CPLX update has also removed it until its potential return

[robust granite]

Can you fix perplexity UI? it's not so intuitive for image generation, selecting specific models etc. Make the UI something like you.com

[urban sierra]

I have a case study that is sort of interesting and causing problems for me. I have a 12 page pdf of 1700s text. It is not OCR'ed so the text information only exists as image data. I am asking Claude Sonnet 3.7 to translate this into digital text, (basically using the AI to do better OCR) which works well except... that when I give Perplexity the document, it only thinks there is between 6-9 pages (even though when viewing the document in the system it still displays all 12) It's also not consistant. I have repeated the process in multiple threads and it thinks the document ends at a different spot each time.

Why would this be and how do I fix it so that it actually has access to read the entire document?

@tropic raft @south ether 👆

[red steeple]

That in Polish literally translates to "Reflecting on the sonnet"

[knotty dagger]

I find the privacy policy of Perplexity AI to be a bit concerning. It says that it holds onto all my searches as long as my account is active, even when I delete it from my history. I just realized that now. It says the only way for it to be deleted within 30 days is to delete my entire account. I’d rather not delete my entire account, but unfortunately might have to if that’s the only way to get my history deleted from the Perplexity AI servers. The other AI companies only retain your searches for 30 days even for active accounts once you delete it from your history.

[dull inlet]

I want to know how I can upload PDF files with more than 800 pages but less than 25MB. Would the Enterprise Pro version have this option?

[silent flame]

how dense is the pdf? any pics? pics can be redcated since the ai can't parse them. If it's all text, it might be better to pull the text into a .txt file

[dull inlet]

Is this pdf

[dusky fjord]

can perplfexity alllow users to use the ai search without requirement for sign up.? I asked my parents to start using it, they liked the first two queries but then couldn't follow through the sign up process. Making it open and searchable without sign up would be amazing imo.

[potent pasture]

We are a company of 17 on the enterprise pro subscription. Is there a management portal where I, the superadmin, can monitor the usage of my team? Y/N?

[compact moat]

who thinks that perplexity is worse than before

[sharp sparrow]

Feedback regarding LaTeX: Perplexity seems to struggle with correctly wrapping LaTeX. It is supposed to use ( ... ) and [ ... ], however for Spaces with some math-related PDFs indexed, it always seems to use $ ... $ which is not rendered as LaTeX. This is happening even with extreme enforcing in the Space's system prompt, and that is, I believe, on top of the enforcement in Perplexity's underlying system prompt. The only thing that is actually getting rendered in LaTeX is $$ ... $$ but the usage of this pattern is generally advised against (and iirc this is also forbidden in the system prompt). The only way to easily mitigate this is by specifying the usage of ( ... ) and [ ... ] in the user prompt.

[fervent nacelle]

Not that I actually managed to get many last time, the easter egg hunt last year sticks out in my memory as a fun community activity.

[tropic raft]

Def one of the things we'll do again.

[green ravine]

Why do you think this is a worthwhile message? You think anyone is getting any benefit from this? Have you searched the server to see if anyone has discussed it before? There are so many junk comments that have absolutely zero value - we don't need any more people just slinging vague and empty negativity.

[fervent nacelle]

I'm curious how a organized event would go on the server, either with the devs showing something off, or for power users to show off their custom spaces/prompts/workflows. I've seen other servers do it and it's usually well received.

[tropic raft]

We're thinking of having various team members show off what they've been working on or doing quick AMAs on what it's like to be at perplexity

still brainstorming so good ideas like the above are ecnouraged

I'd honestly love to have a contest where people show off the coolest thing they've built using PPLX (app and API as separate entries) and let people vote.

I get we're an answer engine but some of yall are using it in ways that defies all logic to me; honestly mindshatteringly creative

[mental matrix]

Similar to how chatGPT is integrated into iOS, having perplexity would be much better. Especially for pro users havu g better information retrieval from the Web would make it really useful.

[hazy barn]

Which variant of the GPT4o does Perplexity use today? Is it the March 26, 2025 version?

[raw minnow]

lmao perplexity giving inaccurate information due to bad sources

the bad source? perplexity themselves 🤣

https://www.perplexity.ai/help-center/en/articles/10354919-what-advanced-ai-models-are-included-in-a-perplexity-pro-subscription

o13 series??

[mental matrix]

I've found badcially always that the more recent the info source the worse perplexity does. Newer sources seems to get drowned out by older ones, so information doesn't get "updated" the new stuff just egts added to the mix, which in an LLM means the pribably of saying the new info is really low unless there is very little old info to compete with.

[raw minnow]

that's a difficult challenge to solve for pplx

for some subjects you would want the latest info

but for others you want to focus on older sources, before all the AI slop hit the internet

[mental matrix]

I think it’s basically fundamental to the way LLM’s work that this problem is unavoidable

[cursive swan]

When researching on data and then asking Perplexity to present the data in a defined tabular format (using as template an earlier output it had perfectly produced), it constantly throws up errors (as can be seen in the attachment)

[twin sinew]

Hi team, when will you add image generation with Chat-GPT 4o ? For the moment we can generate images only with DALL-E, Playground and FLUX

[native tartan]

GPT-4o image generation isn't available on OpenAI's API yet, until that happens adding it isn't possible

[tropic raft]

Think the one right before 4o image gen

[tropic raft]

Can you send me the permalink? Guessing this is web

[dim kite]

I would like to be able to organize the threads in my library by folders, so make it easier to find/categorize. Could that be implemented?

[carmine prairie]

You can make Community on Discourse

[cursive swan]

It is indeed the Web Pro version. Error happens in representing the researched data in tabular format. After redoing it multiple times, it then comes up with the data in tabular format.

[tropic raft]

Feel free to send me a DM of the permalinks where it messed up and I'll flag to the platform and frontend team

likely classifier or some bug was preventing it from rendering correctly

[quaint urchin]

I tried to search a post from a forum.
ChatGPT successfully found it.
But when I switched to perplexity, even using gpt, it can’t provide any result.

Is there any reason?

[tropic raft]

Please provide permalinks (you can DM if sensitive)

[raw minnow]

This new role colour is too similar to regular users. Could you go back to green?

[mental matrix]

Request 10 million for better image generation. So frustrating trying to get good results with tjks strange prompt method

[dull plover]

建議Perplexity 可考慮在未來版本中：
增加「發現」功能內「主題」與「興趣」標籤的自定義選項。
加強用戶語言偏好的整體同步性，確保標籤(科學與技術、藝術與文化、體育、娛樂)與界面語言一致，以提升使用體驗。提供自動翻譯功能

[storm shoal]

The student program doesn't work with my Sciences Po address, despite it being a school in France. What student addresses are actually taken into account

[fringe wedge]

same, hopefully this could be extended to education worldwide.

[tropic raft]

We have a list but clearly not comprehensive. I've informed the team so will let you know once we expand.

[tropic raft]

Is there a specific university you have in mind?

[fringe wedge]

If you have any request process, I’d happily provide an university yes! I can also try to provide a list of domains for a few universities on my country.

[barren brook]

Just for info, Anthropic launched "Claude for Education" today. (But it only covers a selected number of institutions.) https://www.anthropic.com/news/introducing-claude-for-education

[tropic raft]

This should be fixed soon

[sly arrow]

@tropic raft I have identified a major issue https://www.perplexity.ai/search/find-info-on-this-in-practical-R1cO6o0bSuGU49R89VvZ0w
Perplexity can confabulate like crazy. In this case he made up such plausible lie/confabulation as Deep Research - that it is easy to believe in it. It has no sourced.
I am a medical doctor and also researcher, analyzing data, publications etc. I wanted to evaluate and am an avid user of perplexity for quite a while. Actually got 10+ people to use Ppx.
But it is a major problem - it can confabulate data, present n numbers, statistical significance, generate tables etc. Give links to publications (which lead to some chemistry of MnSO4 and something, absolutely unrelated).
I do not use scientific research AT ALL.
It is terrible.
I hope you improve that.

This is a recurring thing sadly. I did tests for specific. Narrow research and confabulated data can be in 40-60% of researches, sometimes completely like that (0 true data)

[tropic raft]

Thanks for letting me know. I'll mention to the answer quality team who can use this to improve the product.

[spring prism]

why the hell is there no redeem voucher option(/link) with no strings attached. Up to a great start😤

https://www.perplexity.ai/pro?discount_code=CODE-HERE
(doesn't actually do anything)

https://www.perplexity.ai/join/p/magenta
(valid code, supposedly invald)

nothing at all works as intended!!

[harsh matrix]

You should spin up a channel for #finance. The domain is big enough to warrant you creating perplexity/finance so probably warrants its own channel here. /finance was better than google's finance out of the box--keep up the momentum.

[sage kelp]

[silent quiver]

No one is responding in Pro Support.
They answered me only on March 28th, since then I have sent quite a lot of messages that have not been answered. It's a little unclear to me if such a poor support job is really "Pro Support".

[covert sun]

Hey @silent quiver! Could you please dm me more details about your case and your email.

[spice dust]

Hello Dear Mods, My college has recieved free access to Perplexity Ai Pro for an year but when i tried claiming it using my college edu email i got pro only for a month while my friends have got it for an year Please Fix this bug

[covert sun]

Hey @spice dust! Please send this report and your account details to [e-mail]. Our support team can check and help with this.

[spice dust]

i did it but no reply from their side till now

[hollow oyster]

I don't know about you guys, but I really would like to have Gemini 2.5 Pro on Perplexity. That model is very powerful.

[tropic raft]

Hey there, unfortunately we can't resolve any student / college related issues here 🫠. Will have to stay tuned from [e-mail].

[tropic raft]

hey @spring prism, thanks for reaching out. I'm not sure what this is about. Could you let me know what issue you're having?

[runic zephyr]

in deep research kindly enable youtube transcript crawl also , this is making perplexity behind its competitors. Thanks for listening to my feedback @tropic raft

[gaunt thorn]

What is the data retention policy related to API calls? I found the Data Retention Policy in the trust center but it doesn't talk about the API side of things. Thanks!

[tropic raft]

For any query data, we do not retain or train on any of that data. For usage data and related meta-data or user account info, we'll store that until you decide to terminate your account + request deletion. Our API ToSo here: https://www.perplexity.ai/hub/legal/perplexity-api-terms-of-service

[charred aspen]

Is there a central repo I can access all my uploaded files? To reuse in queries or just access them?

[tropic raft]

Hey DR! I don't think so unfortunately. They'd be bound to the threads or spaces they're attached to.

[tropic raft]

Definitely. Once it's available, we'll look at how we can incorporate it. Results are quite impressive so far.

[wraith shell]

TTS, the headphone 🎧 is 90% of the time not there. What is the reason for this. I like to hear the natural voice while i read together with it.

Also if it returns a 10 sec backward/forward button would be nice. And maybe a highlight marker so if you look away you can have a clue were we are in reading. But in general (on android app) i don't see the 🎧 on the answers of my searches.

[marsh swift]

Overall I think y'all are doing a great job with perplexity I absolutely love being able to make pages and share them with people. I can deep dive on large overlapping concepts incredibly quickly and have a beautiful cohesive piece with sources and the whole works!

So here are my asks.
1 - Please add the ability to make pages on the mobile app!

2 - please add dictation to the desktop app. This is one of my favorite features about the mobile app and llms like chatgpt that allow you to play from the desktop app.

It's part of my workflow in the day I'll sit down type of everything I need to work through with chat GPT....
And then I'll step into the kitchen which is right next to my desk and do dishes while chat GPT follows up with me audibly.

I would absolutely love to have this feature built into the perplexity desktop app. Truthfully it will probably be even more valuable than the chat GPT one. Because nine times out of 10 I'm getting solid information out of perplexity.

3 - I absolutely love the desktop quick shortcut for perplexity search it's so tidy and spotlight-esque one of my favorite Mac features! And I love having it on be this great place to find accurate information quickly.

4 - I have not for the life of me been able to figure out how to keep the desktop app from fully opening the window at startup. I want it to launch it startup so that I can shortcut key on the keyboard to quickly get perplexity search at my fingertips... But I don't want the whole window to open up every time I turn on my machine.. it's distracting and takes away from my daily workflow. I have my system very dialed in and tailored and this is just a little bit of a adjustment that will make this app feel cleanly built into the system.

I'm excited to see how how you continue to develop your service!
I hope it's built into my Xfinity plan next year because it's literally the only perk that I've ever really enjoyed from Xfinity. 😂

[vague solstice]

I might be missing something but I don't see any UI to create a new page, other than going direct to perplexity.ai/page/new

[vague solstice]

My Library page only has a 'New Thread' button.

Are you a Pro user? If so maybe it's disabled without a paid plan.

[covert sun]

Yes, it's available to Pro users.

[torn ruin]

Is there a way to search the discover pages? i wanted to reference one to someone, but i couldnt see a way to search there so ended up scrolling till i found it.

[gloomy apex]

Not yet, definitely something we are looking to add in the future though

[dawn jolt]

Can you guys add Gemini 2.5 pro now that they have the api released?

[sudden gyro]

Bug in the Windows app: When I start a new thread and enter sth, then switch to another window and come back, the new thread window is gone and the information I entered was lost.

[tropic raft]

@loud carbon

[loud carbon]

@sudden gyro Hi professor, thanks for the report. We are looking into this issue. Will try to report back when we have a fix!

[subtle quartz]

Dear Perplexity Team,

I hope this message finds you well. As a regular user of the Perplexity app, I have a few suggestions to enhance the user experience, particularly on mobile devices.

1. Landscape Mode for Tablets
   Currently, the app does not support landscape mode on tablets. This results in underutilization of screen space, making it less comfortable for extended use. Implementing landscape mode would significantly improve usability and user comfort.

2. Thread Summarization in Library
   When searching through threads in the library, it would be beneficial if the app could summarize the main topic or headline of each thread instead of displaying the original prompt. This would make it easier to quickly identify and revisit relevant conversations.

3. Latest Prompt Display in Threads
   In threads with multiple follow-up questions, the library currently displays the first prompt by default. It would be more convenient if users had the option to view the latest prompt directly, eliminating the need to scroll through the entire conversation.

Thank you for your attention to these suggestions. I look forward to seeing these improvements in future updates.

[fervent spoke]

Hi Perplexity,

I really like using Perplexity, but it's really frustating on some points.

1. I'm a free user, and so I use my free pro searches very selectively. It bugs me that every time I try to ask Perplexity something, I have to manually turn off the pro search, and sometimes it doesn't even care that I toggled it off, it still uses Pro search anyway (like the previous "Auto" search version where there isn't an option to NOT EVER use pro search)

2. When I want the AI to give me a different answer, I don't want to have to edit the prompt or tell it to rewrite, and I expect that clicking the Rewrite button that it will make Perplexity generate a different response! Now even if there was a popup that asks you which model to use to rewrite the response, there should at LEAST be a free option! So the rewrite is completely useless if 1. we don't want to use pro search 2. we ran out of free pro searches! Please make it so we can rewrite a response using the free model!

Thank you.

[solemn pond]

The pro button should default to off.

[whole compass]

To make Perplexity even more versatile, you could add an AI Music Generator. Something like Suno AI or similar?

[whole compass]

No, it shouldn't. As a Pro User, this would be dramatic. Maybe only on the Free Tier.

[fluid escarp]

https://www.perplexity.ai/search/que-es-una-parte-electronica-n-40av9NJPQeOg2.F9rH1eJA wth I asked to identify an electronic part and it gave me references to Colombian medical something and other completely unrelated subjects. Search results have gotten the last few weeks really bad.

[latent otter]

Will you add manus ai to your model repository?

[gentle stirrup]

I'm a PRO user, since couple hours I am facing troubles generating images in decent quality, because you suddenly and without any previous announcement: you downgraded the quality to a shi*y 512x512, is it a temporary bug or a new (bad) decision?

[tropic raft]

Hey Error, thanks for your feedback. We'll defintely include it in our meeting with the product team re: how we can improve our user experience!

We ask that you refrain from "." or "bump" in these threads, as they crowd the chat for other users. We read these messages and will keep you posted if these changes are incoming.

[tropic raft]

Hey there, manus isn't a model so not sure how we could include it?

[tropic raft]

Thanks for letting us know Lazy Boy! We'll make sure to include it as key feedback.

[fervent spoke]

okay 👍
sorry

[stoic island]

Hey, everytime I go to perplexity.ai it always defaults to pro search - how do I change the default behavior?

[silent quiver]

done!

[tropic raft]

Hey sirgoodwill, the "pro-best" will route your query to the best model capable of answering your question. Is there a specific reason you want to change it?

[proven imp]

can't get any of the old threads , each time perplexity updates it ruins the current thread and it go's like this

[crisp fjord]

Is Perplexity’s podcast “Discover Daily” still going? Last ep was in March.

[sly vector]

For some reason, when responding to my prompt, part of the prompt response was in an entirely different language. Any idea?

Reference link: https://www.perplexity.ai/search/what-does-gemini-2-0-flash-hav-8YuyaHroTqiIsP2.CMzauA

[final trout]

Has this happened before?

[pure granite]

Hi @everyone

I am using perplexity instead of Google search the recent upgrade pro and deep search when to use where to use I am bit confused it provides unrelated answers too while searching can you pls guide me

[mighty crystal]

Yes, O2 has done the same promotion (free subscription for a year)

[final trout]

Talking to support. If there's a fix, I'll share it. Otherwise, email [e-mail] to push the issue

[sly vector]

Prompt engineering is a thing that any AI model needs. You need to make sure you are asking in a way that will give the answer you want. The broader you are, the more likely the answer will be somewhat unrelated. This isn’t just perplexity but other LLMs.

[pure granite]

Previously it was searching for documentation providing the correct answer now it provides anonymous answers

[sly vector]

If the information is not sensitive, can you share a link to your perplexity thread?

[sinful rover]

Can we please get the ability for sticky sources? Like when I select Web and Social, it stays like that for future queries.

It's small, but annoying after time when I have to do it for every query

[pure granite]

Is it fine which is related to technical

[sly vector]

Sure!

[pure granite]

@mragentshar how to share the link

I am unable to see the options

[fiery shard]

guys can you add mcp support to windows app

[frigid latch]

Hey, can perplexity add a go to the bottom of a thread button in macos and ios version of the app like the windows version does. for some reason my command + arrow down is not working 😦 i need to scroll all the way down everytime 😦

[torn ruin]

thank you

[tawdry vapor]

What happened to the ability to choose a model on iOS? This option has disappeared from settings.

I still see it in the Android app.

[heady plover]

Perplexity Pro user here, I used to be able to select "Reasoning" and "R1" as my one and only model to answer questions, now the toggle reverts to "Auto" not just each time I start a conversation, but even within the same conversation, for every message I have to reselect "Reasoning". This is annoying af. If you struggle to pay the bills don't cheap out like this, I want to be able to select my preferred model and it has to stick. Raise prices if you're struggling

[daring valve]

Using Perplexity for the last 3 months, would love if the file size could be bigger then 25MB. I understand the larger costs, but even like one model that has the increased file size would be amazing

[tropic raft]

hey Arena, you can select Pro > one of the reasoing models (o3, R1) to have reasoning on by default. LMK if that doesn't work.

[heady plover]

It works temporarily, for one prompt. It doesn't stick even for one conversation, and certainly not across conversations. This is on iOS, latest version of the app.

[digital garden]

Please make it possible to delete specific sources from threads. A simple example: I created a thread related to nutrition, and at one point, I uploaded a photo of my meal, asking about its balance, etc. And literally, every subsequent message (regardless of how many) references this photo. However, it’s impossible to delete it since users can no longer remove the sources they’ve added. Hoping for your understanding, from an active Perplexity user.

[heady plover]

This is obviously cost cutting. Create a higher tier if necessary but don't be cheap by default

All I am asking is a dropdown to select the model which is consistent across the various platforms and it sticks, at least within the same conversation. You've been messing with the model selection UI every other release and it gets more cumbersome and annoying each time.

[swift night]

This has been happening to me too on the mobile app, but I haven't tried it on web recently. I actually needed to reinstall just for the app to register that I had a subscription at all, but then the mode would always toggle back to Auto whatever I did.

[heady plover]

It is of course intentional. The whole model selector logic has been a battleground for the last 20 releases or so, it is constantly fighting my legitimate desire to have R1 selected always invariably no matter what because I say so. Again happy to pay a higher price if they could stop f****** with my model selection

[keen osprey]

Please remove the system prompts bc what even is this

[slim cobalt]

Hi. I'm nor sure if this topic, correspond with this channel, but I don't know exactly where to write it.
Is there any information or roadmap about the future of the Perplexity Assitant on Android. I'm amazed on how it works right now, but I really hope some kind of integration with my email, calendar, notes, Telegram, WhatsApp, etc.
I've searched but I haven't find any relevant information. Thanks! And great work!

[torpid geyser]

Is there anyway to turn the search function on and off? It helps when I need it but it interferes when I don't and running two instances of perplexity so I can use it selectively and copy-pasting between the two is just painfull.

[hard edge]

Is search down? Using android app none of my queries use any online sources/search no matter what focus I use

[queen merlin]

Perplexity Pro cannot be convinced that Biden did not win the election, and that Trump is President. Here's the bookmark:

https://www.npr.org/2025/01/20/g-s1-43759/trump-inaugural-address-key-moments

Sorry, this one. The other is one the the links I provided in my effort to convince it.

https://www.perplexity.ai/search/provide-the-complete-list-of-n-1GfX.uKyS2Wa35b92tHTYw

[hard arrow]

In space pplx unable to read files uploaded as context for any models.

[zealous zenith]

Please expand Perplexity Finance into global stock market, not only US, because it's really helpful

[boreal lotus]

Just learned after bug reporting that the option to swap a prompt from a Space to another has been deprecated because it was buggy, one more reason to stop subscription for me... Going from love to hate with this really

Of course it still possible on Android because there's so much incoherencies between web and apps, dont know if i should be happy about this or not

[vast pecan]

I hope I post this in the right subject, but I need an explanation. In a conversation I provided an XML file to the ai and asked to make changes and deliver me the resultfile back. It seems that he does this, but the resultfile is not downloadable. If I click on it, I get an access denied. Isn't it possible to download resultfiles fromo a conversation? (using pro version)

[frail crypt]

add also FX in Perplexity Fin

[polar heath]

A suggestion - Add the functionality of converting files in "spaces" and "internal knowledge bases" to markdown using Mistrals OCR model before you put them into the vector database. Retrival accuracy and readability of complex PDFs - that are designed for human eyes - seems to be MUCH better when you do this markdown conversion with mistrals model.

Alternatively, just let users upload markdown files themselves manually which then get vectored. Please at least to this!

Please let me know your thoughts.

[heady plover]

As a Perplexity Pro member, I’m disappointed with the inconsistency of photos and images since the latest update. Previously, there was always an option to generate photos or videos if they didn’t show up automatically, but that option no longer exists. Sometimes, for queries where media would be helpful, no images or videos show up at all, and the experience has become very inconsistent. Media also makes the answers more visually appealing and engaging.

[lyric nimbus]

if it isnt too long, you can ask it to put resulting text into a codeblock in response. but pplx has a bit confusing system prompt, the report writing llm (synthesis) often forgets a user doesn't necessarily see code execution outputs. access denied might be either pplx migrating the file storage, or the model hallucinating (depending where you see the file in ui)
example: https://www.perplexity.ai/search/pretend-you-are-a-professional-fsInC4JDTmeAqpqMWin6Dw

[lyric nimbus]

slightly confused - by generate, do you mean show image/video search, those tiles now response tabs; or more like "Generate image" where some ai creates a whole new image?
I think probably that search image/video tiles. it sometimes doesnt give anything, but usually you can just ask and quickly get a bunch of videos
I agree it often made the responses nicer, but there was a lot of cases when those images were irrelevant and only distracting. also new ui is probably better suited for phones, and I may not like it, but if you have same layout for desktop and phone, it gets a lot easier to maintain and extend

[turbid glacier]

any idea on real time results...I was a little surprised when PPXY told me it can't do real time and "My knowledge cutoff is set to April 2023"?? as in two years ago?

[heady plover]

yeah, i meant the ability to search for images. While it can be distracting, just having the ability to tap the image section to see it is better than having nothing. It's been very inconsistent - sometimes for certain people's names or places it doesn't show up, and sometimes it does show up.

[lyric nimbus]

well, not entirely realtime (sometimes it fails to see day or two old news), but if you keep enabled web search in sources, it should be able to get pretty recent events. for example here: https://www.perplexity.ai/search/any-news-about-openai-today-dv4ib6GSQvCEctF092gOnQ at least the memory update is true and very recent.
actually I am surprised any llm on pplx said to you its cutoff date, I though they had this forbidden in sys prompt 😅

[valid hinge]

I didn't know the Bulls are now competing in the NFL lol

[turbid glacier]

I know....and to just offer that date up like that.....haha

[tawdry vapor]

The weird thing is that the model selector is still available on Android but gone from iOS

Can someone from Perplexity comment on this? I guess I'll reach out to customer support

[crimson nymph]

You can select the model in the top right on iOS when typing your question

[tawdry vapor]

Thanks. Although it's bottom left actually.

Also, the options are these. Not how it used to be and still is on Android like ChatGPT, Claude, etc.

[crimson nymph]

That is the mode, you can select the model in the top right

[tawdry vapor]

Weird. I don't have that.

I'll also check on ipad

[tropic raft]

Please go to "Pro" first to select model in the top right. "Auto" will simply select the model best fit for the question, which is why you can't choose a specific model.

[tawdry vapor]

Thank you! 👏

[junior eagle]

How to disable PRO searches when I dont want them? It automatically upgrades queries to PRO, many times on simple questions, and using PRO search allotments without consent. Very frustrating.

[wooden creek]

What model do you have selected? I think the 'Best' option is akin to the old 'Auto' option.

[swift night]

A few people have mentioned earlier in this server that we can't change the model any more. It just defaults to Auto or switches back whatever we do. Mine is still like this (Android). Is it being looked into...?

[lyric nimbus]

I dont use the app much. just tried (android) and while I cannot select specific model for a new thread, it uses sonnet which I think I had selected when it was removed from settings. new pro thread and even response are written by sonnet (I didn't see a model [that chip/cpu icon] mentioned in responses; but checked on web).
this is just a speculation: so maybe you have to switch it on the web version and maybe it will stick for the app? if you end up trying that, dont forget to close (ideally that forced close) the app, then switch it in the web version on home page, maybe do one search with that model on web pplx, and then try to open the app
ps: shouldn't this q be rather in #📱︱feedback-mobile ?

[junior eagle]

I dont see a way to choose a model anymore. If I goto Settings | Preferences .. Choose Model, it just redirects to homepage. You have option of PRO, DEEP RESEARCH. I toggled both of them OFF. But Perplexity "upgrades" my searches to PRO whenever it wants to, even for simple queries .. using up PRO search allotments for silly stuff. I'm hoping there's a way to stop this.

[swift night]

I made a post in there already but it received no response, I thought that maybe since other people had already mentioned that issue in this channel it might be more compelling for someone to answer.

Thanks, I'm happy to try that, but by saying you can't change model in Android either surely this is worth noting as a bug? Further, I don't think it's a very good feature asking users to switch between versions just to change a single setting, so realistically this is feedback across both Android and Web.

@lyric nimbus and as you can see from the reply above mine it really is affecting quite a few users at this point.

[lyric nimbus]

yes, definitively can be considered a bug. but with apps pplx may have already prepared the release and it may took few days before google/apple gives a green light. on web they can roll out changes almost immediately

[swift night]

I'm happy to try to actually write it up as a proper bug report but I sort of suspect it ought to be on the radar by now with so many people talking about it.

I mean I'm not really sure what additional detail it would realistically require, but I can try when I get home.

[neon depot]

Why Perplexity Student Referral not recognising the .edu.... E-mails

[lyric nimbus]

yeah, definitely check #1140622008086970420 , if it exists give the + or thumbs up, or add details. it can be a bit of a pain filling that on phone, but has a bit higher chance of devs seeing it, if poeple keep bumping it with updates

[swift night]

I'll try to have a look when I get home. Thanks for your replies!

[lyric nimbus]

I vaguely remember they had 50% off for edu mails. though I am not sure if those edu accounts are supposed to count as referrals. also werent there something like free pplx for month or year if school got enough accounts? I suspect that was only for US.
it is possible I got something mixed up, I think chatgpt also has something for students now

[covert sun]

Hey @swift night! Which app version are you using?

[swift night]

Hi there, I believe the app version is 2.43.0. My device is running Android 14.

[covert sun]

You still have 'Focus' in the top right corner, when you're sending a query, right?

[covert sun]

In this case, you can find the Pro model switcher in your settings. Could you please check that it's answered with that chosen model and when you want to start a new query you can start it with Pro?

[junior eagle]

How to disable PRO searches when I dont want them? It automatically upgrades queries to PRO, many times on simple questions, and using PRO search allotments without consent. Very frustrating. I dont see a way to choose a model anymore. If I goto Settings | Preferences .. Choose Model, it just redirects to homepage. You have option of PRO, DEEP RESEARCH. I toggled both of them OFF. But Perplexity "upgrades" my searches to PRO whenever it wants to, even for simple queries .. using up PRO search allotments for silly stuff. I'm hoping there's a way to stop this.

[covert sun]

Hey @junior eagle! Is the main reason you're reporting this that Pro feels slow when the questions are simple and don’t need many steps? Have you tried the Sonar model and the "Best" option - does it still feel like Pro overcomplicating things with those too?

[heady plover]

The model resets to "Pro > Sonar" or "Auto" not only when I start a new conversation, but for EACH PROMPT even within the same conversation. Let's not pretend this is a bizarre occurrence, it's obviously cost cutting.

[junior eagle]

2 reasons actually .. I dont want to use PRO on simple queries (or unimportant ones), when PRO options are limited. Second, is yes, it takes a long time to answer a simple question. I dont see where "BEST" is even a choice anymore .. just toggle PRO or DEEP RESEARCH .. Which I untoggled, but Perplexity "upgrades" to pro when it feels like it.

[heady plover]

What I want, and probably everyone really wants, is that I select my preferred mode and goddamn model and it will f** stick. How hard it is? Can I pay you money extra for that?

[covert sun]

You can set the model in this menu.

[covert sun]

Hey @heady plover! What version are you using? This has been changed a while ago, where models don't switch to auto anymore. And the Pro model you choose is used for follow-ups.

[junior eagle]

I dont see that menu. Is that because I need to upgrade?

[heady plover]

iOS, 2.43.0 build 9190.
Not only do I have to select "Reasoning > R1" for each and every new conversation, but - and this makes my blood boil - I have to reselect it within the very same conversation for each follow up.

I get used to a certain model and learn to make best use of it, until at some point I move on to the next. Do NOT change my choice unless I say so.

[covert sun]

That's in the Pro model list, correct.

[obtuse dawn]

Why doesn't the Gemini 2.0 Flash Experimental option appear in this menu?

[hard arrow]

same for me

[lyric nimbus]

well, I can see it, but ...

I asked for an image and it looks like it generated it using python? I don't think this is "hyper-detailed render" with "cinematic lighting" 🫠

for a mathplot it is good, but not really what asked for 😅

[heady plover]

there should be a arrow button to see nxt image on same page, i've to click each generation to see it

and if I go to images section, previews there are smaller

[heady plover]

Arrow buttons are on the bottom once you click on "Generated Image"

[heady plover]

i get it, it will be really efficient if that option comes on main page

[little jewel]

Boy do i have feedback

1. Its unclear wether youre getting the model you have selected to answer yoir question. On multiple ocasions i have gotten answers i was sure were gemerated by a different model to the one i selected.

2. Its incredibly unclear what is happening after you press search. Some vague bullet points are generated by a different model, and they dont at all represent what is actually happening.

3. Reasoning models reasoning is split into bullet points for no reason whatsoever, to make it look like there is more going on than there actually is. I would much rarther be told something along the lines of "wait for it to do its thing", than be lied to about what its actually doing.

4. The context window of models is cut wayyy down. And the user is never informed of this anywhere.

5. The system prompt for perplexity features its entire wiki pasted in for no reason, increesing cost for perplexity, degrading service quality, and... i cant think of a single good thing that comes from this

6. The model doesnt know what model it is. You paste in the entire wiki and time down to the minute but you dont tell it what model it is

So from all that, you get a service where you might not be getting your selected model, you dont know what its doing while you wait, the ui is just unresonable, the model doesnt even know what you said 5 messages ago, the answers are much worse than the base models.

How come?
Why for?

[heady plover]

1. That's because they constantly reset the model to "Pro > Sonar" because it saves them money is what it is...

They are sneaky af about it and there is no reason, they could transparently charge extra for a tier where you can always use a reasoning model you selected , without let or hindrance

I would pay that, with glee

[pallid charm]

I keep getting these issues when asking questions that require heavy mathematical notation. Anybody know how to fix this?

[lyric nimbus]

1. I see this a lot on reddit, but I havent seen any solid proof yet. I remember with R1, many people lamenting it is a cheap knockoff (distill), yet when I tested it,it was reliably consistently answering questions which only proper R1 did. even the biggest distill had like 10% success rate while big official R1 and pplx R1 had like 90% success rate

2. why do you think they are not real? system reads question,decides what to do next (search->reading; repeat; or skip) and it's given to report writing llm which user selected (in case of reasoners their thinking is appended to pipeline output)

3. again, why do you think it is a lie? did you try comparing sonnet 3.7 thinking with claudeai? or 1776 with official R1? I remember trying the latter and didnt really find anything signalling a different model or being a "lie"

4. that is on the user. they should either ask on discord or reddit (if they are lazy), or find the answer on the discord or pplx webpage (see images). also it is quite comical, because perplexity is probably most open about its limits from all big names. chatgpt, claudeai, le chat, all these has some vague x times of free account or more than free account and never exactly stated what exact limits free account has. pplx has officially 300+ daily smaller responses of 32k ctx window. in reality it is 600 response daily (500 reasoning), with file up 62k token window and some reasoners have 128k context

5. that was fixed, pplx boss already addressed it on reddit. yeah, not great, it was a bit annoying for just few days, but most claims about it were out of proportions. it was 15k characters = ~3k tokens, so no, it didnt eat half of tokens of smaller context window of 32k/62k/128k

6. they could, but if they swap name, it would be terrible pr. safer to say all models they are Perplexity. many people forget pplx is for search, not chat. thats why it has bias to interpret next query as standalone one. tell me, where I get sonnet 3.7 think 32k+ for 20$/m?

[lyric nimbus]

wait, that was also fixed some time ago. last few days I think it happened to me just once after deep research. or does it still do this in apps? (I use virtually only web)

6. I mean hundreds of uses daily

also in 4. I forgot grok. those I think have exact numbers stated

[heady plover]

There is nothing to "fix" because it's obviously and clearly a cost cutting exercise. It's not a bug, it's a feature and if we push against it maybe they will backtrack on this nonsense.

[little jewel]

1. For me personally, i think that the reasoning models arent faked, because thats simply harder to do, but i was testing gemini 2.5 pro on perplexity, and i am damn sure they were giving me a worse model

2. I gave it a pdf with 30 physics questions, and told it to solve all 30. It prints bullet points:
   Solve task 1
   Solve task 2
   Solve task 3
   And so on up to 30
   It hangs on the first point for about 30 seconds, then the next 29 points are instantly completed.
   So clearly the bullet point progression doesnt represent what is happening.
   Also, i give it a task like repeat this text back to me, and it makes a bullet point that says "repeating the given text back to the user" hangs on that for like 10 seconds and only then proceeds to do it. Another example of the bullet point saying something that isnt actually happening.

3. I didnt explain myself properly there. What i mean is, when you ask a question, it first does some searching, then it says begining analisys, and lets the model think for a few seconds before answering, and that thinking, the way that it is displayed on the client side, it is split into these chunks with the same ui as the research bullet points, but its being split for no reason, and would be more resonable to print the whole reasoning chain as 1 continuous block. This is a client side display only issiue. It doesnt affect model performance.

4. Im not taking about usage limits. In that i do agree that perplexity is more transparent than others. I am talking about context windows of the models - how many tokens of previous conversation is allowed to be crammed into the model before perplexity cuts it off. And in this regard, perplexity is much less transparent than others. The other companies just give you as much context as the model supports, but perplexity just cuts it off at an arbitrary length, and said length is not disclosed to a user before paying for pro. And that is borderline scam behaviour, because a user could within reason assume

That perexity allows for usage of the full context window of the available models, ya know, like every other company

[little jewel]

5. I tested it today and it still injected the long ass prompt, hopefully they fix it soon tho

6. Youre not getting 32k reasoning tokens lol. The whole ass responce length (reasoning + message) is hard capped at 4k. But where can you get claude reasoning for $20 a month? From the official claude site, t3chat, you dot com, and probably a few more.
   Also these providers dont cut context windows without disclosure

[little jewel]

I dont believe a normal ass person should have to go to some discord server to seek out basic information about the product they are purchasing

[little jewel]

Slay
File size hard capped at 32k tokens for a model that supports 1048k tokens

[lyric nimbus]

they announced they are no longer pre-selecting auto after original query, and that is mostly working on the web #📢│announcements message like I said, I got only few times "Best" from deep research, but never from normal (pro) search or space. and I am not a light user

[heady plover]

On the iOS app they haven't changed it

[lyric nimbus]

1. I personally didnt spent enough time using the new gemini. in cursor it is pretty solid, though for me, I dont know if it is really better than sonnet 3.7. btw technically, the gemini 2.5 pro is reasoning model (but reasoning tokens are not returned via api, not even summarization like openai does)

2. that sounds to me like a RAG problem. and yes, perplexity has its own RAG, filahandling. to doesnt neccessarily say anything about the model. I think pply uses only vision from models which support it. pretty sure saw few bug reports about models no seeing uploaded files. could be related. what you see, the steps, thats pipeline/workflow/process, that I believe doesnt handle selected model by a user.

3. sure, for thinking they could do it (I think originally it maybe was?) without cool bullets, but thats only a cosmetic thing. also sometimes the planner constructs longer plan than needed, so it can do like 1 point, then decide it has enough data, so it skips few next points. that is not a bug though, more like feature (mostly to cut cost, but it is true that models can start giving worse results if overwhelmed). also some ui stuff is still buggy, I still see broken misaligned text during search phases. they iterate too fast, so there is always some bugs. usually nothing major stays for too long

4. I partly agree, it is not documented exact context handling. but I am pretty sure they always allocate same space for searches (which are not left in history of thread for a last model to see), final synthesis model only sees last searches.

perplexity is much less transparent than others
hard to say, I am pretty sure claudeai doesnt give full 200k window to free users. and I think it was similar to chatgpt
The other companies just give you as much context as the model supports,
I am pretty sure this is not generally true. and there are tradeoff, just today I saw angry claudeai user telling he quits because he could only do 5 messages. just different tradeoffs

[little jewel]

Fair enough

Cant get fairer than fair enough

[lyric nimbus]

I tested it today and it still injected the long ass prompt, hopefully they fix it soon tho
it should be fixed, meaning it will get injected only when a user asks about pplx, or at least better than before where it was injected like every second thread. I was trying it today and even directly asking about perplexity and my sub status didnt trigger it, soo 🤷‍♂️ maybe a/b?
Youre not getting 32k reasoning tokens lol. The whole ass responce length (reasoning + message) is hard capped at 4k. But where can you get claude reasoning for $20 a month? From the official claude site, t3chat, you dot com, and probably a few more.
I meant only small thinking, even 1k thinking can move results quite drastically (and I don't think pplx has much over few k thinking). just did a basic math and in api it would cost me 2x more than perplexity and that was sober 50 uses, 4k in, 1k out and no search api. sure, I didnt check its math, but doesnt seem unrealistic. and there are days I do few hundreds, though part is usually non-thinking
I dont believe a normal ass person should have to go to some discord server to seek out basic information about the product they are purchasing
it is also on perplexity page...
and yes, normal person should check what they are paying for. personally I find non having guaranteed uses like chatgpt does behaviour which should be illegal. how many images daily has now chatgtp? oh, nobody knows, because it is "dynamic". claudeai launched 200$ plan and suddenly 20$ witch 5x more than free account is quarter of its size what was last month.
pplx at least has some guranteed stuff, like that 300+ pro uses, 32k+ context window
Also these providers dont cut context windows without disclosure
yes, they do. few months back free sonnet 3.6 didnt allow file upload around 150k

[lyric nimbus]

it is a different product than native platform has. also it is 62k tokens, ale least for sonnet 3.7. on native platform is also different web search, file handling and code execution. so why context window should be any different?

[lyric nimbus]

it is possible. not only pplx has issues with this, because even if they upload app release same moment they upgrade web, it still waits for apple/google approval before it gets to users, days, maybe a week 😕

and just to be clear, I am not saying perplexity is perfect and better than any other service. no, it all is about user's use cases. if you want upload your whole codebase and work on it with ai, than better pick would be cluadeai or even gemini, I guess. didnt expect 2.5 pro to be so good at programming. similarly if you primarily want ai images, then something like midjourney or ideogram (I hope I didnt butcher the name) is better. heck on pplx it was for years so hidden the image gen, than there wasnt a day somebody on discord didnt ask about it. just recently, like yesterday thay added it as a tool, so you can trigger it from normal query. but for, lets say light to lower medium length search, pplx is great. and despite many saying it is not suitable for chat, it is not bad if you disable web search, or do some light anchoring every few messages. for more serious search and news, grok might be better. for long researches chatgpt.
pplx has a lot of weird stuff, but is pretty well rounded - a lot of bigger models, though usually not biggest ones (opus50, I am sorry, I still miss it). rag in search is okay if you know how it works (~128k characters from start of file), similarly spaces (instructions cannot affect pipeline, so eg search or code execution, and RAG is useless for some programming languages). but others have similarly weird hidden limitations, often dynamic ones which I just hate.
ps: I documented many more limits -> https://monnef.gitlab.io/by-ai/2025/pplx-tech-props so definitelly not a pplx shill 😅

[barren brook]

" and suddenly 20$ witch 5x more than free account is quarter of its size what was last month."
????
I’ve been an Anthropic customer for a long time, and since the launch of Claude’s two new subscription plans (Max and 5-20) this Wednesday, I haven’t noticed any change or downgrade to the Pro offering. My experience remains exactly the same as before.

[lyric nimbus]

I dont know if it is true, but saw a bunch of posts on reddit https://www.reddit.com/r/ClaudeAI/comments/1jwqukz/just_5_messages_in_pro_before_hitting_the_limit/ https://www.reddit.com/r/ClaudeAI/comments/1jx4p73/is_anthropic_rate_limiting_pro_users_much_more/ https://www.reddit.com/r/ClaudeAI/comments/1jx8qp0/with_max_rolled_out_pro_limits_just_went_to_the/ , plus people regret paying pro for a year and they change the servic so drastically it is no longer usable for them

[barren brook]

I’m not Reddit, I’m a real user. Believe whoever you want, but as far as I’m concerned, my experience hasn’t changed at all since Wednesday—the day the new subscriptions launched. 😉

[amber ocean]

Still inside actual claude platform, the chat says to start a new chat 😂 due to conext limit cross 😂, so basically an intelligent AI but not of use.

[barren brook]

With the Pro Plan ??

[amber ocean]

Now im not using actual claude platform anymore. But my friend who is using 100usd plan is still getting that issue😂

[lyric nimbus]

I saw many times reports on Anthropic having troubles with lack of compute, sometimes even with API (like in Cursor), not only claudeai the platform

it might be more factors, like time and region. I think they first start limiting their platform than api users

[amber ocean]

They are with a concept of if context window crossed then start a new chat, thats the issue, chatgpt tackles it by sending latest amount of context as input rather than asking to close chat, claude developers are unfortunately not interested to implement this concept 😂

[barren brook]

You can find criticism for anything if you really want to 😅 Anyway, I was just sharing my perspective as a real user — I’ll leave it at that.😉

[heady plover]

It is possible because I am observing it right now, on a release of the app which is literally 4 day old.
They are just cheapskates hoping I wouldn't notice.

[vale topaz]

hello, any plans to bring dark mode to the Chrome extension?

[dire drift]

Feature Suggestion: add swipe to Delete option

Why: it takes too many clicks to delete an entry

Platform: MacOS desktop

Extra Notes: Each search creates a separate entry. At the end of the day there too many entries so it would be ideal to have the option to delete them quickly.

[wet wasp]

It's better in maths but still off when it comes to code

[amber ocean]

Anyone using gemini workspace? Let me know how good it is?

[pale widget]

Unless I am misunderstanding this we should not be getting charged for pro searches if it was upgraded automatically

[covert sun]

Thanks for reporting, we'll check.

[fading storm]

i specifically de-select pro for my search
pro search is used anyway
immediately run out of free pro searches on crap that didn't need it

WHY

this is genuinely getting on my nerves, i keep coming back around to perplexity even after saying i'm not using it again, because it's in my workflow and it's what i use daily now

but what the hell is this whole idea of pushing pro onto people when they don't want to use it

like okay, sure i'm not paying for perplexity (and at this rate, i don't think i ever will), but if you're gonna dictate what searches do and don't have pro search then why bother offering free pro searches in the first place????

[celest axle]

I think that if perplexity could make images and edit those images would be better

[native tartan]

Good news: https://www.perplexity.ai/changelog/april-11th-2025-product-update

[pale widget]

This genuinely pisses me off as well. They need to add an option to disable it entirely

I have a feeling they’re pushing it for some sort of data collection purposes

[tame grotto]

Make product offerings clearer. Separaring reasoning model and other models got confusing and cluttered. Gemini 2.5 Pro being in the settings model selection, not in the reasoning model dropdoen is horrendous. No consistency at all.

[amber ocean]

Hi,

Is perplexity good for long chats as GPT4.1 1 million context is available in Perplexity?

[lyric nimbus]

it has some consistency - new gemini in speed is more closer to non-thinking models and since thinking tokens are not sent from API, for pplx it very well might be closer to non-thinking models.
in browser version it is already in one menu, will probably take some time to get it into apps

[lyric nimbus]

no. just tried it and it failed after 7th caption, so that is 27k tokens. mentioned (incorrectly numbered) caption on 58k, so probably entirely same file handling (in search, not spaces) as all other models

since we never got the 1 million context window from gemini flash which I think is cheaper (didnt check, might be wrong), I wouldn't expect 1 million from 4.1 😕

[amber ocean]

Yeah one guy (user) responded, gpt4.1 is 128000 in Perplexity. Not sure if it really works for atleast 128000

[lyric nimbus]

that might be for sources from web, or could be mistake - perplexity has some limits in characters (exact same every time) not tokens (+- with each attempt).
thought 128k tokens was for reasoners. upload limit in search is approx 128k characters (~62k tokens), same as other models

[peak sapphire]

For the perplexity referral program, a lot of my friends signed up using my referral link but I didn’t get one month Pro subscription for any of them. Who can help here?

[nimble latch]

Perplexity I know you guys have to save money but changing the selected model back to auto is annoying
either way I check every time and every single time I switch to the model I want
so please disable the annoyTheCustomerToSave2Bucks setting for me ty

"We do this to ensure every user gets to use the models" yeah just tell me you need to save money, heard that 1 million times already on every single subscription based AI tool ever

[copper dagger]

We need more than 32k token though… please consider it.

[fluid quarry]

Hi, this is Shogo.
We’re currently developing an app that uses Perplexity Search via the Dify platform.
I have a question: Are user search queries entered into Perplexity Search ever used for AI model training?
I checked the official documentation but couldn’t find a clear answer on this.
Are data inputs sent through Perplexity Search used for AI training?
If so, how exactly are they utilized?
I’d appreciate any clarification you can provide. Thanks!

[full shale]

Is anyone else seriously frustrated with how bad Deep Research has gotten?

Just two weeks ago, the CEO was tweeting like, “We’re gonna make it amazing in a couple days—anyone got a better name than Deep Research High?”
Now that option is gone, and honestly, Deep Research is worse than the original version.

Back then, they hyped it up like it was going to blow Manus out of the water—coding support, external tools, the whole deal.
Meanwhile, it feels like every other AI company has a better research tool, regardless of the model they use.

What happened?

[barren kayak]

Everyone in my family loves perplexity and we all are using it as our go-to for many of our searching needs and we would like to use pro for some searches, but our needs do not justify an enterprise plan. How about a family plan for 4 named users, with limits like the individual plan, for a modest price? $30/month $300/year?

[grim birch]

1. Voice recitation of answers on PC, like the one on phone
2. Adding variety of sports, like chess, and more finance options, like EGX, actually, some kind of integration with TradingView would be amazing, it's another market for you guys!
3. Hire me as intern
   Until we meet again 👋🏼💀🏃🏼‍♂️‍➡️

[timber coyote]

Replying to my own post about using up Pro tokens when Pro is not needed. Solution: don't sign in. This will use vanilla P.a only.

[covert sun]

Hey Janex! Could you please share the latest threads that were worse than the earlier ones. We'll check.

[covert sun]

Hey @nimble latch! What is the app version that you are using? This had been updated and the pro searches you have shouldn't switch to auto, there is no 'auto' in the latest model list updates, too.

[nimble latch]

I think it's called "Best" now

And it's the web app

I am using Gemini 2.5 and the search defaults to "Best" from time to time

[tropic raft]

Hi Shogo - the data is not stored so can’t be used for training. We only store metadata like usage volume and time stamps for support reasons.

[fluid quarry]

@tropic raft
Thank you for your response.
We understand that the data is not stored and will not be used for training purposes.

Is there any mention in the official website or terms of use that data is not cached and will not be used for training purposes?

[viral shuttle]

Please enable o3 and o4-mini right now. I need something better than Gemini 2.5 pro.

[kindred epoch]

What Does This Mean Plz

[full shale]

please keep this website up to date: "https://www.perplexity.ai/help-center/en/articles/10352901-what-is-perplexity-pro"

You stopped support for 4.5
You added 4.1 and Gemini 2.5 Pro

Also I do not care about grok, but their api released for grok 3 and its certainly better then grok 2 (which makes it extremely redundant.)

[opaque echo]

In a regular query, I can attach c and h files just fine, but I can't upload c and h files in a space, it says unsupported files. Is this by design or is it a bug?

[lyric nimbus]

grok 2 was pretty bad. grok 1 released as open-weight was a waste of space - much smaller models were way smarter and even the output didnt seem too good, I think only at writing at some niches it was somewhat interesting.
their "open" releases are comicly terrible, just pr gimmick so thay can say the released (some uselss garbage) as open-weight. grok 1 wasnt useful at all, anywhere near sota (even relative, compared to size/cost/speed), nowhere near mistral or meta "open" releases, even worse than openai and that says something when openai didnt release anything useful in llm space for like a decade, but at least whisper and embeds were pretty good and I believe were sota for quite a long time (half a year? maybe even more). and grok 2 is most likely too large to be of any use now, so virtually same fail as grok 1, I mean in open-weight category.
but grok 3 looks quite good. I agree xAI are a scum company - they did train on benchmarks and manipulated their measurements (the whole cons debacle and faked bar heights), but the model itself seems quite solid. the deep search on top of it on x and grok com hallucinates much less than pplx "research" (what used to be named "deep research") and is fairly competent at programming. I actutally hope the "deep research high" from pplx will be at least of quality of grok deep search (hallucinations and length; writing already is pretty good).
though xAI is like a company of extremes. they are massive liars when it comes to hype-building and benches, but on the other hand - they are like only company which isnt selling "we can deliver zero uses of all our services per month and fullfil the contract, and you stupid user will happily pay $$$ for it". chatgpt, claudeai, le chat (mistral) - they all are selling "more x times free tier; but no exact data on what free tier has guaranteed" bullshit. xAI is only company which clearly states probably all limits on buy page. pplx has public some limits, but far from all

[hollow horizon]

Chatgpt's plus limits are available in the FAQ section

[covert sun]

Hey @opaque echo! Could you please dm me the file you are trying to upload, we'll check if something is broken.

[zinc vector]

I want to add my voice to those requesting a setting that will disable the Newsfeed that currently appears below the Search Bar on the Home and New Thread pages. I don't want to be distracted by click-bait headlines about politics, Elon Musk, or anything else. I don't want trends or suggestions unless I ask for them. This is not helpful or benign, it is distracting and potentially very irritating. One of the great strength's of Perplexity for me is the beautifully simple and elegant interface that lets me focus on my work, without interference from the outside world. Please give me an option to keep it that way.

[solid mauve]

Don't know if this was suggested before, but:

You know how in the ChatGPT app or the DeepSeek app, when you regenerate a response, you can go back to the older generated response by clicking an arrow, and you can cycle through responses, and you can actually create branches this way. I would like to suggest adding this feature with perplexity. It makes things a lot more organized, because the way it is right now, if you go back and regenerate a response from a previous point in time, things get really confusing.

[wet wasp]

Regarding the voice mode, I find that the model speaks way too slowly and I can tell it to speak faster, but every time I do it has to start from the beginning of its response. Would be great to have the ability for it to be preset to speak faster

[oak crest]

https://www.reddit.com/r/perplexity_ai/comments/1g69okn/stop_perplexity_pro_from_referencing_previously/

Do Perplexity has any intention to solve this issue? Is it a bug or Its a feature ? If it is a feature, how does it help? I find it very irritating

[misty kestrel]

Why chatting with gpt models in perplexity is less accurate than actually chatting with them on chatgpt? Is there any way to fix that?

[barren brook]

PPLX uses OpenAI’s API. The system prompt (instructions), the context window size, and even the effort parameter (if applicable) can differ significantly from what you're used to on ChatGPT.
This can affect how the model behaves when you're using it on PPLX.

Perplexity is, above all, a search engine — but a high-performance one, built for speed and precision.
Every task has its ideal tool — it’s up to you to choose the one that fits best.

[lyric nimbus]

that is still worse than grok, because they have it right in the plan. also isnt it hidden somewhere? perplexity didnt even found faq main page and from chatgpt, I had to click on a tiny question mark at bottom right to get to faq section, searched for "limit" and didnt found aynthing relevant, most was for API. do they really have those limits as exact numbers, like for free you are getting 5 requests to GPT-4.1 every 5 hours and dalle3 uses 2/5h, 4o image gen 2/5h etc?

[outer moss]

macos Hover input box can't paste picture, it can in the internal interface of the software.

[fading storm]

this is what i can't understand.

perplexity auto-enables pro search for free users so their free tier is used up
perplexity pro users can't keep their selection and end up using the free model

it's more than frustrating

it's not even like it's a "saving money" thing (because the first wouldn't happen), it's just anti-consumer at this point

[crude skiff]

Can we get candle-stick charts in Finance?

[heady plover]

I have been shouting about this from the rooftops for weeks now, they need to end this bullshit and do it yesterday or else I am stopping my subscription and influencing as many people as I can not to purchase Pro

[restive jungle]

man the image generation feature is really frustrating

how is it possible that refining my prompt continues to produce worse and worse results

if i set a custom image prompt its reset every single time

[kindred shard]

Hi, this is Francisco from Spain! I am surprised to see the app has no spanish voice recognition for Spanish speakers. I present volunteer if you need any help with that!

[hollow horizon]

https://www.perplexity.ai/search/solve-cFcS4bHaTmWBIo.QAJlALA

https://www.perplexity.ai/search/solve-wc1YAOhNSnO9Fw76mAcXQw

messed up outputs

[vivid cave]

Hello, I just wanted to appreciate the new event yall are doing for students, thanks to you guys, I can learn better. Will love perplexity because of this goated stuff.

[heady plover]

Hello perplexity team, could you add back a UI In library that told us shared thread or private thread?
So I don't have to check It manually.

[marsh zephyr]

Need side scroll bar on Perplexity Mac app. It gets so hard when we have long conversations

[fallow pier]

need feature to track indian stocks under Personalize>Finance section. Rightnow only US stocks are available to be added to watchlist.

[celest axle]

I would add if it’s possible a document displayer for the AI because I’m trying to generate a PDF or a Doc and it just says it can’t send me that type of document file

[lyric nimbus]

yeah, only images are supported as output from code execution ("Programming" step). there could be few ways around it

• use AIlin and run code for generating pdf/doc in a docker container (safe, but available only on pc and web version, harder to setup)
• try some alternative way how to encode the documents you want into an image format (if you are not a programmer then don't try; probably even professionals will hit some roadblocks)
• let the ai write it in markdown or html and manually then convert it using some tool to pdf (pandoc? probably will be some free online services)
• instruct the ai to as a last step in the code execution to upload a result to somewhere (you would probably need to share api key or password, a bit unsafe, since you can easily leak this info if you ever share a thread with this used)

[night palm]

@covert sun when will the “check sources” feature work on the iPad or iPhone? It doesn’t even work in the browser version on iPad or iPhone, but the option shows up on the iPad

[heady plover]

Follow ups in image gen dosen't work ig, i hope it can happen

[lyric nimbus]

it can work. but even first query in thread has severe limitations. I have used prompt template from here: #1355669590096482495 message for both - the initial message and the followup

[rocky oriole]

I think it would be nice to have the modes back, writing and math and etc… but this time, it should include image gen so you don’t have to specify in the prompt

Other than that, I like the accuracy!

[languid sandal]

Anyone having problems with the chrome extension not working?

When trying to use the extension, it shows this error. The web application works perfectly fine. Might be a bug or something.

[grand holly]

Lemme try

[grand holly]

Yep. Not working.

[languid sandal]

Maybe something with the API or connection between the extension and application, as the website works fine. Thank you for confirming is not only me! 🙏

[wanton mulch]

Misleading results - I was just asking for “thought experiments” and digging deeper by requesting examples. One suggestion was for thought experiments in Social Science, Psychology and SocialPsychology gave me examples of the unethical research that led to the development of IRBs.

[grand holly]

What's irbs?

[opal snow]

Hi, anybody has a problem with an any private link sometime?

[covert sun]

Thanks for reporting, @languid sandal, we'll check

[covert sun]

Hey @opal snow! What kind of problem do you have with such links?

[supple summit]

Hi! I read an article recently about how Wikipedia and other open-source resources are struggling under the weight of LLM queries, and I noticed at the end of a recent session that Perplexity heavily utilized Wikipedia as an information source. I've been considering becoming a paying customer of Perplexity, but I would want to know that this invaluable public information resource is being protected. It would make me a lot less hesitant to pay for the service if I knew that some of the proceeds went to support Wikipedia. I know I can donate personally, but what I'm really hoping is that it would apply to all paying customers. Have you considered incorporating that into the pricing as a way to be socially proactive and simultaneously enhance your brand's public image? It might even be possible to make some sort of arrangement with them to give your service priority over other, non-contributing ones.

(Awesome product, btw!)

[zinc vector]

I am confused by the changes to Spaces in the April 18 release. It appears I can no longer assign a Thread to a Space by clicking the three dots next to the Thread while in Libary view. My only options are bookmark or delete. (I thought bookmarks no longer exist, and in any case I can't find them, so why is this still an option?). It seems I now have to go inside a Thread in order to assign it to a Space. This requires more steps and is much less convenient than it was before. Am I missing something?

[short zodiac]

Trying to sign to the API overview event, but website is not sending confirmation code. I’m using a relay email from Apple (as in [e-mail])

[lyric nimbus]

a lot of people doesn't see wikipedia in positive light (heavy left bias). I personally would prefer to blacklist wikipeadia from my queries (along many news sites)
also I am pretty sure they aggressively are caching most popular wikipedia pages, because it is good for pplx - faster responses. possibly even cloning whole wikipedia and using it as an internal "db". though both of these approaches have a slight trade off - cached wikipedia pages might not be too current, so AI would "see" older version

[vagrant moss]

one thing i'm noticing is that some of my spaces are becoming crowded and could use some kind of sub-space or folder-like structure to them.

[orchid copper]

I started a support chain over a week ago, got one response, and then nothing over multiple follow-ups. I would just like my current enterprise organization to be deleted

[covert sun]

Hey @orchid copper! There's an option to delete your organization, and you can do it on the web version. There was an issue with that button that has since been fixed.

If you're still experiencing problems, please let me know. I'll need your account details to check with the enterprise team.

[orchid copper]

This is what I see on my org page when viewing from chrome on my mac, should I be looking elsewhere?

[covert sun]

Yeah, there should be the Delete button. Let me check

[supple summit]

Weird that a publicly editable information site that can be corrected by anyone if it's less than factual would be considered left-biased. Seems like an information bubble, IMO, if that and various news sites are excluded. Diversity of information sources is for the mind as diversity in gut microbiota is for the digestive system -- protective.

[covert sun]

Hey @orchid copper! Could you please clear cache and let me know if you see the delete button on another browser?

[empty crystal]

can we control the search time range?
tried with a prompt instruction but it didn't seem to work

[fleet sleet]

@orchid copper I am not able to download the invoices. Can you please help me?

[sick helm]

Can we add custom emoji (or image) support to spaces icons?

[lyric nimbus]

I saw a lot of agenda pushing on wikipedia. majority of pollitical and related topics are based on opinion pieces of legacy news media. that few center/right wing media is usually banned on wikipedia, while ignoring same offenses from left leaning ones. after all even the founder of wikipedia is aware of it. the problem is it has a tendency creating bubbles - leftists value their time much less for a free work and more of them is either unemployed and calling for ubi or something or they work for scraps just to push their ideology (very similar thing with reddit)
if you look at left/right bias of news organizations and schools (from primary to universities), it is very biased toward left, in some cases even extreme left.

[supple summit]

I'm a "leftist" by some definitions and it looks the opposite from where I am standing. This is the divide that has been created in US politics by social media and filter bubbles. IDK how to undo it without someone sacrificing their principles. I think we should probably step back from this conversation before we stray into full on political discussion and violate the rules here. Hoping for an eventual resolution, even if none is in sight right now, which is to say, best wishes.

[lyric nimbus]

my point was - pplx should not by default funnel money from subscribers to other services, especially if those services/entities are recognized as strongly political. in this case probably even an opposite of what CEO's stance seems to be
not sure if it is even needed (see the caching), but if implemented, it would have to be opt-in, up to a user to set, eg send a cent from each search which uses given service. hard to tell if even that isn't too much, since you can ask in pro search to get extra sources, you can get over 100. so probably more fair would be tracking if the source was used in the answer in a substantial way (eg more than few words) and user selected explicitly splitting money. of course then would have to be implemented a blacklist, because I think a lot of people even now don't want some sites to be searched in (I think it used to support this feature at one time, in an ai profile; now I think only in queries themselves, though it is so cumbersome I just rather say in ai profile which sources, eg to ignore "professional" gaming critics like ign, so the last AI handles the filtering/ignoring), and if some donation would be forced, that could bring a lot of backlash
to be honest, I always though about games and movies that it should be the other way around - after finishing a game, leaving a cinema or turning tv off, then the customer should decide if and how much to pay, possibly with suggestions from the author/publisher/dev.
I kinda like the idea of money splitting eg humble bundle. a user has few sliders, deciding how much goes to which party, so if they dont like one game in a pack, they just set 0% for a publisher/dev of that game (so more goes to other parties). maybe similar thing could be done on pplx, but probably on a monthly basis with remember option?

[mild kernel]

A button to directly go to the last query in a long thread would be a lifesaver.

[frank bolt]

☝🏻 hopefully waiting. I have some very long threads and to go to bottom the easiest way is to type hi or random word, then it goes down. But I think it burns company money and button would help so much

[barren brook]

Thanks to the PPLX team for hearing us out on this issue. 😉👍
1)Thread

2)Space

[lyric nimbus]

isnt it already there? or do you something else?

[mental matrix]

Wikipedia doesn’t have a bias. Facts have a left wing bias.

[heady plover]

Image gen selector just like model selector on home will make things easy coz i have to switch between gpt and dlle everytime by going in settings

[lethal flint]

Will voice mode support text prompts at some point?🤔

[supple summit]

Makes sense. Maybe just a default of, some of the money goes to the sources, whatever they are, and the user has control over which sources are used. My primary concern is that, in a post-search world where the LLM does the sourcing for us, revenue for these information sources from direct traffic is going to go into decline, while cost is going to increase, so we are going to have to come up with a different economic model to ensure that primary information sources stay in business. Otherwise, primary sources will start evaporating, and then what will these LLMs even have to search for us?

[mental matrix]

it'll start searching AI sources, which will eventually encompass all sources, with LLM's recursively being trained on and searching other AI created text. That would be a fucking disaster.

[unkempt glacier]

It would be great if it were possible to disable news tiles on Perplexity.

I’m trying to be more mindful about my news consumption. Perplexity’s insistence on featuring U.S. political news prominently is particularly unhelpful in this regard.

Perplexity considering news sources in answers is fine. It’s the fact that news headlines are shown when I don’t want to see them that I have issue with.

[little jewel]

Context window short as fuck
Please fix

[opaque echo]

Thank you for rolling out the new GPT image generation so quickly. It's a great addition, and I really appreciate how fast the team implemented it after the API becoming available. One thing I miss is a dedicated photo or image button, so I don’t have to type “create image of…” every time I want to generate a picture.

Also, when I try to modify a previously generated image with a prompt like “give that guy glasses,” the system doesn’t seem to understand which image or person I’m referring to.

I made a space with the instruction "create an image of my prompt" but that also did not work hahaha

[neat vigil]

I concur with this sentiment. It would be highly advantageous to augment the existing ‘Search’ and ‘Research’ functions with supplementary buttons, specifically incorporating distinct ones for ‘Image’ and ‘Chat’. The ‘Image’ button would be dedicated to facilitating effortless image generation, while the ‘Chat’ button would operate under a default setting that disables web access as a source, thereby permitting its function purely as a conventional chatbot.

[fervent nacelle]

Could we clean up the #💗│sharingchannel and remove all the image posts? it was nice having that channel as shared threads/pages only.

[barren brook]

I believe this is a temporary solution, while waiting for an image channel to be available. 😉

Ah, I just saw that it exists. 😅

[subtle latch]

I’d love to see a feature that allows me to preview the voice options instead of having a discussion over and over to test out all the different voices.

[true scroll]

Is there a reason the iOS app used to work fantastic when asking perplexity a question compared to today when it loads forever, and never answers the question and forgets the question was ever asked?

[opaque echo]

yea agree, the gemini app implements this wonderfully

[true scroll]

I would love for it to work

👆

They’re also needs to be a better way to get in touch with technical support for assistance.

👆

[mild kernel]

@lyric nimbus NOT on mobile.

[lyric nimbus]

oh, I see. since you brought the attention to this button, I found it is not that great on web either. It will show when you reload or navigate to a thread, but if you scroll just a bit down, the button disappears and I didn't found a way to get it back. so it could use some love on all platforms it seems 😅

[mild kernel]

And trust me the mobile ui became horrendous lately. Do you know it stopped showing me newline? Entire prompts for stories I am doing without newline, without enter button. All are on same one line space. I am using gboard on moto g60. Just a flick and all disappears. I have often lost several minutes, ten to fifteen minutes of work for a small flick of a finger.

[lyric nimbus]

I don't use the mobile app much. But I remember losing few of my prompts this way. and that feels like months back 😢

[supple summit]

The lack of newlines is something I was frustrated with recently, too. Other than that, I'm fairly pleased with the mobile interface

[torn marsh]

I was using Gemini before coming here. Figure why use just Gemini when I can use this and have more options. However the one thing that I really liked about Gemini is the Canvas feature. That is such a useful way to interact with the AI, makes it feel more colaborative and I find it easier to follow along with things like if I'm working on some code or something. Would love to see a feature like that here.

[dreamy swallow]

My would-have-been 5 day streak for 1 month of pro was reset! I am an obsessive all day/everyday/everything user of Perplexity without needing to be challenged... but I was so anxious to be able to use pro again for my OCPD. Sorry for sounding like a freeloader, but how can I avoid my streak being wiped again?

[mild kernel]

@supple summit Is there a workaround?

[supple summit]

No, I just awkwardly slide the text to the left or right. At least that works, so I'm able to read it, even if it doesn't wrap like I'd prefer.

[vagrant moss]

I could not possibly be less excited about the idea of another browser that tracks my every move.

[signal atlas]

Same. It's a horrible idea and I'm outta here. Hope your company gets ruined, Perplexity.

[restive ocean]

Hi 👋 Please stop having the app (and maybe web?) change my search defaults. Every time I search I have to make sure it didn’t turn off Pro or switch it from GPT to sonar.

[next knoll]

Seems to be a major bug in spaces search on iOS with the latest app that came out yesterday. Upvote here if you're affected. https://discord.com/channels/1047197230748151888/1365257889253556224

[fervent nacelle]

We need reasoning models in the iOS rewrite menu.

[clever vine]

@covert sun
Please add this data for chats that the AIs use for colors, such as Gemini 2.5 Pro. For example: <font color="purple">Example text."</font>
Apply it for mobiles and the official website, thank you.

[clever vine]

@covert sun
Gpt 4.1 displays the following function for text colors:
<span style="color:Teal"><b>Example text:</b></span>

[vocal vortex]

#👉│feedback-general
Attachment linking while writing prompt:

It would be good to be able to link specific files to my prompt

imagine having space with 25 pdfs and links listed

and you want to manually link pdfs to the prompt how Notion does that already
this would be great

[nimble latch]

Please stop switching to "Best" automatically

[vocal ore]

iOS version unusable :feelsbadman:

[digital garden]

It seems that the new image generator from OpenAI, integrated into Perplexity, performs poorly or doesn’t work at all with languages other than English. When attempting to request image generation in another language (in my case, Russian), the model states that it cannot generate images and suggests alternatives in the form of other services or applications. However, when using English with prompts starting with "Generate an image of..." or "Create an image of...," image generation works seamlessly.
Additionally, it’s worth mentioning that if the preferred language in Perplexity’s settings is set to something other than English, and the image generation request deviates from the standard phrasing like "Generate an image of" or "Create an image of"-for example, using "Make a picture of" or something similar-then, in most cases, the model (responding in the language set in Perplexity’s settings) also claims it cannot generate an image and offers alternative solutions.

Examples provided on images
UPD: and looks like image generation not working with unticked Web "source for search"

[polar comet]

There's currently no way to export a Page to PDF or MD correct?

[halcyon granite]

hey, loving perplexity pro. wanted to provide feedback which may be an edge case.

im a native english-speaker living in japan.
sometimes i feed it a japanese webpage or two and ask - in english - for explanation, summary, etc, and it responds entirely in japanese. so every time this happens, i need to follow up or ensure my initial query says '<prompt>, reply in english'.

i guess i initially expected that it would respond to me in the same language as my prompt rather than the data. it would be great if this was the default behavior, imo!

[night palm]

I have a Mac that runs macOS 12.7, and it can not update safari, and perplexity.ai no longer works on my computer's browser

@ok.alex

[naive grove]

Hey, im new here. Thought I’d start with some feature requests: Could the discover feed deliver country specific news and have more detailed curation options. Couple examples. I’m based in the UK and just see US news which is great but I’d like to see more news about my government too. Would be great to calibrate this somehow. Second: I’m a big anime fan and have an ova launch I’m really looking forward to but hardly get any anime news in the feed would be great to more granularly configure it so you can see these sorts of topical interests. Also if I were say the team creating that anime, I’d love to set up a page that can give anime fans the ability to “discover” my anime and follow it early possibly even pay to advertise it on the platform. (If you’re wondering what the anime is it’s Shinsei galverse ☺️) Discover is by far my most used perplexity feature. I love it. Can’t wait to see how it evolves

[vivid shell]

Plexicity team , i 2nd this, I want to see more localized based on user locsation and should be optional, as not everyone is US based and the ouputs plexity gives me might assume or relating only to u.s.a market which I am.not in not in.

Plexisity thought I'm united states based "based on IP" which i am no were near

[shell forge]

Ability to increase text size inside settings to make it bigger.

[burnt viper]

I second this. Should be location flexibility

[late heart]

HI, I'm a heavy user of Spaces on Perplexity Pro. I appreciate that Perplexity is now able to choose the best model for a specific query, but I actively use the other features provided by spaces, including special instructions and designated sources etc. The current inferface does not have a method to easily access space via a search box. Instead I have to scroll through the spaces. Please create a search box for spaces, or let me know if I am missing something about the interface. Thank you.

[sacred tide]

There used to be an images button I could tap to search for images after I’ve already done my query. Is it coming back?

[shut arch]

3. Treat community members with respect.

No harassment, hate speech, or malicious behavior. Keep things positive and constructive.

[crimson nymph]

please keep this channel related to feedback of the product

[bleak helm]

Stop removing model choice!!!!! Perplexity has the choice of being my best app (before) and the app I hate most (now) !!!!
The ongoing changes and differences among the model choice are maximal annoying ! On my iPad now it allows only the choice between best/sonar, this is NOT acceptable. I will abandon Perplexity, if this will going on. I love very much the exact opposite where I can choose between the models, as currently in my iPhone and my Mac. Stop treating us as children!! If you want to make the UI easier for normal people, do it, but let a "real Pro" option available !!!!!!!!!!!

Generally, the ever ongoing UI changes, especially between the different hardware models (e.g. web, Mac, iPad, iPhone) are not acceptible!!! Why you are making a good app unusable and to hate??? If you are unable to provide a consistent interface between the apps, just use the web app which can be used nearly as native apps for everything ! I don't accept, that every second day you remove some model choice for some of your apps !

[bleak helm]

General rule for your underperforming UI QA : STOP REMOVING FEATURES because you want to simplify UI ! Instead, allow the choice between new and old behaviour with a switch (at least in the Pro version). I would give the UI variation a keyword as ("simplified 1, only choice Best/Sonar, no ind. model choice, 4/24/2025"

Suggestion to be more precise on LLM names: Add more information (high/mipublishing model date to the LLM model names. Especially the old name "latest GPT-4o" was incorrect. As OpenAI and others change models regularly, the "last change or published date makes a difference. Additionally o4-mini could be attributed with "normal" or "midi".

Feedback on model choice here: This is very important for "Pros" as me.
I LOVE that GPT-4.1, Gemini 2.5 Pro and o4-mini were included so fast! And that we have two more choices in Reasoning than only Deepseek R1.
On the other hand, I would love to see even more expensive models available. You could add "Premium" price models, where this is possible. At least I want to try them out. I would appreciate a "Premium fair" price model where I would pay some fee for EVERY question on premium models like O3, especially O3 deep research, if I just I want to play with them. It would allow me not to have multiple subscriptions outside Perplexity !

[lyric nimbus]

o4-mini could be attributed with "normal" or "midi".
at least in the model picker, I don't think so. staff said it selects dynamically on its own if it uses medium or high. 😩 but in answers it could be nice to see what o4-mini it was. personally, I would like version split in the model selection, high could have stricter limit (eg 100 per day or something)

similar thing with GPT Image 1 - after some number of images, which they didnt disclose, based on some undisclosed algo they degrade img1 performance so much, it is worse than 2 year old local models which people were/are running on their midrange pcs

I wish they would be more transparent - state all limits, never hide model names which were used in response creation during whole pipeline/workflow. you can have by default "Auto" for everything. but for more advanced users, there should be an option to see what is going on (eg model names, better error messages) and be able to select what they want (eg o4-mini-high, image ratio, img1 high quality [and to see how much is remaining], temperature)

[mental matrix]

Deep research is especially useful for generating longer form responses to outdated ideologies that have been disproven by science. I kinda think it should be turned off after the first response, my follow ups are basically never deep research worthy.

[bleak helm]

@Kirai: I agree with you. Moreover, every dynamical select is misleading for some professional users who are also interested in LLM model differences and not only answers! At least, If model choice (e.g. low/medium/high) is dynamic, write so! I dislike every underlying dynamic. I want to have the option to compare model answers. If there is a problem with the price model, then create a price model with limits for dedicated use of "o4 mini-high" or GPT Image instead of degredating. Silent degredating is not for professionals. Just create a "premium price plan" where you are told if you run out of the normal pro limit and let us pay for the more questions or choose cheaper models !

I think we need a new reddit message of Aravind, the UI confusion has become to much the last weeks, you just effectively steal and rebuild model choice every second day !
But one positive feedback: Taking away the reasoning choice is ok for me. I liked the original, but just choosing the model e.g. 4-mini is ok for me. (But again: "Stealing" model choice at all from my iPad Perplexity make me furious!)

[stable flame]

Please keep Grok API access, it is so good with the Perplexity spaces, especially in following the prompts, best across all other models!!! Appreciated!!!!!

[coarse quarry]

I've been getting friends and family to use perplexity bc it's awesome. Then when the promotion for a free month started a few weeks ago I used that to convince them to get it and use it. I was already using it every day so stacking up 7 days in a row seemed like automatic. However I've now filled the card out twice and both times it just reset even with 20 days of consecutive searches, across devices . Why does it fill up the reset and not fulfill its promotion ? Am I missing some instruction ...what am I doing wrong

[pastel scroll]

at the beginning or when deep research high was temporarily released yes
but not atm

[grim birch]

I know people exploiting promocode programs, selling a year's subscription for a month-worth, does it matter for you guys? Should I report them somehow or subscribe with them? 😂

[lyric nimbus]

careful with those, I saw few users in "general" complaining about the missing pro or account access and I think they lost it because pplx found out about some reselling. not sure after how long, but if you start using pro, it might sting a bit losing search/chat history, spaces etc

[grim birch]

Actually I'm bounty-hunting, maybe they'll give me a legit annual subscription if I give them the promocode 😂

[lyric nimbus]

GPT Image instead of degredating. Silent degredating is not for professionals.
it might not be intentional, at least that looks to me from a few responses on discord
At least, If model choice (e.g. low/medium/high) is dynamic, write so! I dislike every underlying dynamic. I want to have the option to compare model answers.
yep, especially if it looks like a user is getting X model, it should be at least communicated it is X model plus some secret tech (custom feature name, sparkles icon with tooltip etc). I dont expect them to "open" eg Research mode, but there at least nobody is expecting one specific model.
when I was comparing offerings of AI platforms, it felt a bit strange finding out Grok has the most transparent limits, directly in a subscription plan selection, clearly how many uses and refill time. claudeai and chatgpt has nonexact vague phrases like "more usage" and "extended limits", compared to free variants which lack any concrete numbers. so, xAI which cheated/mislead several times with Grok 3 (charts - "cons" and bar heights, from what we know they probably trained on benches to get artificially better scores) - and this company is the most transparent one?

[jolly cedar]

I’d like to be able to disable the news features in Perplexity. I already stopped using the macOS app in favor of the web app so I can block the news elements. I just deleted the mobile app because there’s no way to disable news there. It’s just too distracting

[unreal edge]

i would like to disable showing the app in the dock option from the app settings

[jolly cedar]

You mean showing the app icon in the bottom dock on macOS or are you talking about having it in the top bar?

[unreal edge]

i don't want it to be visible in the dock (bottom) but visible in the menubar (top bar) it doesn't have to be a turned on by default I want it to be able to toggle it from settings because not every one will like this feature

[jolly cedar]

Yeah I don’t think macOS would allow for an app to be open that’s not visible in the dock. I could be wrong but I don’t think I’ve ever seen that. I think they should add the option to not show it in the top bar/not let it run in the background though

[unreal edge]

there is a lot of apps that have this as a toggle feature chatGPT used to have it but they removed it recently

[jolly cedar]

Ah ok never mind then

[jolly cedar]

The biggest reason I like AI-assisted web searching is that it helps me stay focused instead of getting lost on random sites, but with the news feature it completely messes that up. I’m switching back to ChatGPT until i find something better

[haughty ledge]

Hi, I wanted to know why iOS and macOS apps do not contain "Introduce yourself". I find this very useful as the system prompt, currently it works only on the web but this field is there since forever. Is this on the roadmap to implement?

[lyric nimbus]

yeah, AI profile is experiencing a rough period. I'll start with a good thing: if you set "Introduce Yourself" on the web version, it should work on mobile too (it is saved on Perplexity's server)

on web:

• misses "additional questions" section, but they are still passed to models. which means a user has no way, for example, to remove some custom formatting or remove old hobby etc

in apps:

• you cannot change introduction nor additional questions

Is this on the roadmap to implement?
apps in general are lagging a lot in features. I think image generation isnt still there? recently there were some new bugs, like prompt box only for one line, search not working in spaces

[digital garden]

Addition: It seems that in (my case) Russian, the generator works 100 out of 100 times only with the starting phrase "Нарисуй" (Draw). Phrases like "Сгенерируй" (Generate) or "Создай" (Create) either don't work or almost never work. I hope for a quick fix! 🙂 @covert sun

[foggy plaza]

Perplexity should add an option for free users to toggle PRO mode usage, I use perplexity as a default search engine and I am a free user. The problem is when i try to search something it automatically uses my pro queries without me specifically choosing it
Maybe an option in settings to disable it?

[narrow wyvern]

Same here. I’ve detailed that in my report. Is this a deliberate thing then?

Because the whole time I thought it was either a bug or just a problem on my end.

[foggy plaza]

I think so It's been happening multiple times. Sometimes I don't even notice the queries being used, and when i really need to get deep research or better results i don't have any queries left.

I try to stop the response but its already used up

not sure if this a tactic to make us buy pro or not lol

[narrow wyvern]

Yeah, it’s kind of annoying. Means I’m constantly at 0 queries. Other things I’ve noticed are that every chat with web searches I’ve made has the Pro icon next to it, and that if you switch off web search, it will do so at some point regardless.

Don’t know if all that is supposed to happen or if the recent updates to the system have caused other things like Pro and web search to break and come undone.

Reported it a day ago for clarity, but nothing’s come back.

[foggy plaza]

Yeah exactly, it does that to basic searches like sometimes I ask it to convert usd to rupees and stuff.
I'll make a report as well it's very annoying as a student myself, since i utilize deep research a lot

[narrow wyvern]

I also use it for my studies. Not sure what part of your academic journey you’re at, but I use it for my GCSEs, which are swiftly approaching.

I’m either using Deep Research at times for case studies, or using non-web search mode and ask it to give me practice questions for English Language and English Literature papers.

Trouble is one’s siphoned by automatic Pro search and the other is temperamental.

Great to see you’ve reported it as well. Let’s hope this reaches someone.

[lyric nimbus]

wait, this used to be a search param. but I think it is broken for some time. ill try to dig up the thread

[lyric nimbus]

did you try modifying search url to https://www.perplexity.ai/search?q=%s&copilot=false? I have pro sub, but when I used it like https://www.perplexity.ai/search?q=test&copilot=false it seems to be using "Auto" "Best" in that thread instead

[foggy plaza]

copilot=false isn't a valid param recognized by perplexity, so it doesn't do anything just gets ignored
I think the problem is "Auto" tries to guess if the search needs Pro and turns it on. This is a problem because most of the time since it just enables pro for every basic quick searches

[lyric nimbus]

yeah, looks like you are right. copilot=true maybe does something, but copilot=false probably never worked.
https://www.perplexity.ai/search/test-58NYSOg_To6Fyy2FyqN64w
kinda comedic, perplexity should be "knowledge engine", but they don't document what their search url supports...

[toxic nova]

I need the shortcut to work system-wide on my Mac, whether the app is running or not (this works perfectly fine for ChatGPT)

As it stands, the shortcut to start an inquiry with the LLM does not auto-launch the app or the input box (preferred)

So I have Perplexity as part of my log-in items, so it's ready if I hit the shortcut key. I don't want a splash page or anything else to show up when rebooting my Mac

Is there a way to either do this? Or better yet, to allow the shortcut to auto-launch Perplexity's input box whether or not the app is currently running?

In other words, the same behavior as the ChatGPT app if possible..

thanks

[lavish elk]

What happened to image generation? The tab is completely gone and I can't say anything that'll get it to generate an image. 😲

[crimson current]

I'm not even sure where to begin with image generation anymore.

[tribal garden]

In WhatsApp perplexity, I’m seeing a big empty space

[marble jetty]

I just signed up to Pro because Perplexity told me that if I did I could reference previous conversations for context (to avoid having to repeat myself), but then it couldn’t do that.

Then I asked “well what can you do” and it told me it was “faster and more detailed” so I said “I don’t care about fast or detailed. What features does Pro give me that Free doesn’t?” And it said “we can maybe do some extra things. Let me know what you want to do and I’ll see if I can do it!” And I said “how am I supposed to know what to ask you to do? Can’t you just tell me what you can do and then I can figure out how to use it?” And no. No it can’t. lol why did I get Pro? I even made a Space, put the previous threads in it for context AND submitted the links to the threads for context, but it couldn’t tell me any details about the reference threads.

I MUST be missing something. All I want is to be able to continue a conversation from day to day using the voice functionality.

[stable flame]

Can anyone add the ability to rename the threads in spaces please, it will be really helpful

[barren brook]

Open the thread in question, click on its title, and you’ll be able to edit it.

[severe pewter]

Suggestion: Graph view feature like Obsidian to link content in Spaces.

Suggestion: Focus mode which does not have distracting suggested articles below the search box.

[severe pewter]

Agreed. It does often lose context. If a follow-up question does not explicitly link to the previous question, it can often revert to generic responses.

[lavish elk]

I've figured a temporary workaround. If you search for something that returns results with pictures, you will be given an option to generate a picture.

[stable flame]

but it is still not changed in the spaces (web), i assume it only changes the download document?

[barren brook]

Please refresh the page...

[barren brook]

For example:

[mystic frost]

Hello @shut arch I'm loving Perplexity so far.

[mint frost]

Hi perplexity. I can’t upload files or paste long queries when at work because my employer blocks access to Amazon S3. Would be nice if the file could be uploaded to perplexity’s server instead of directly to S3 from the client. Maybe an advanced option?

[zenith mantle]

Hi! Is there a way to zoom in/zoom out in desktop application (windows 11) or make font larger? It would be useful feature for me since I'm using it in browser with 120% zoom

[lilac musk]

Is image generation coming back to the iPhone app, it was working, now it’s been “turned off”

[opal snow]

Hi, Alex. Sorry for the late and thank for your availability. I have a link about a chat that wrote sometime ago. I work on another computer and I see an alert about authorization from owner of this link. But I'm sure that I'm the owner of this link and I don't have some notification, What could be?
Thank you again for your time, cheers.

[brazen vortex]

So image generation thru WhatsApp has been working for awhile and now there’s no response at all?

[eager elm]

How can I turn off push notifications related to “news” or “trends”? Very annoying to have yet another app telling me this stuff and most of the time I’m wondering why are you telling me X when Y is of global importance.

The Arts and Culture segment is also broken or your curators are bad. Top stories in Arts and Culture: Nike & NFTs, Starbucks 3D Prints a Store, Japan buying foreign rice, LAX Metro

[lunar garden]

I cannot generate images with the pro version. Can someone help me with this?

[halcyon condor]

I use Perplexity Pro both with iPhone and PC (with application).
In the last period PC application asks to re-login very frequently which is quite boring since I have to use Apple ID 2-factor authentication.
Any suggestion to avoid it?

[karmic sluice]

I have one big gripe with perplexity almost never providing links to the official site of whatever I'm inquiring about unless asked for specifically. It boggles the mind that this is not default behavior, and that there is no way in the settings to instruct it to do that. It seems to me the first thing a search engine should remember is to be a search engine. Which means it should find things. That should be core functionality. It smoothens the user experience so much. Do I have to fire up Google every time I need a direct link?

[heady plover]

Hey, could you please share some cases where navigational search did not work? Thanks

[karmic sluice]

I am not sure I understand what "navigational search" means, but I will be more clear about what I meant. Funny, you know how something pissed you off a dozen times but can't remember any one example? :))

An imaginary example is "what are the best AI Writing apps? I have these requirements [blah blah list requirements], which one is the better fit?" My experience has been that it gives a nice thorough answer, listing several products and comparing the offering based on my requirements, and will list the sources, such as some articles about those products, but what it won't often provide are links to the homepages of those apps. So I have to ask for them specifically, waste time, burn more tokens and get irritated.

I feel like if you're giving me a list of products for example, it's a no-brainer that each one should have a link to the homepage. It's a quality of life thing not to have to copy-paste the results into google.

EDIT: Here, I found an example
https://www.perplexity.ai/search/what-s-a-good-screen-recorder-g1TLNjf_RfS8RirPrBGTwA?0=d

[heady plover]

https://www.perplexity.ai/search/is-grok-3-5-going-to-be-releas-prEjrnu6TDKEu_keYJIVsg

This answer has incorrect info due to a possible SEO spam site https://malline.id/technology/2025/04/4285/grok-3-5-vs-grok-3-ai-comparison/

I recommend Perplexity to stop using this website as a source.

[open palm]

A very small feedback, not as important as others but, i feel there could be some picture for the gmail account of perplexity supply, just looks professional for companies to have company logo for all their emails. ✌️ ✌️

[little jewel]

Perplexity will do anything instead of increesing the context window

[wooden creek]

It might be interesting if there was an option to use a higher context window, but it would count as more than one Pro Search or something?

Just don't use the 'tokens' or 'credits' approach as other sites, those are mostly just to obfuscate the cost.

[blazing copper]

god bless who ever re-designed the webUI, reduces decision fatigue by alot ❤️

[sonic seal]

I generally like the new WebUI. However, it's not intuitive to display the library and bookmarks from the Home button. I'd like to have a Library button to the sidebar.

[heady plover]

Thanks for reporting!

[last crystal]

what did PP change with pro earch and best model?

[nimble rain]

A chevron-style expand and collapse would be very useful. Because sometimes there are hundreds of searches on the same page and it becomes extremely difficult to find a search and you need to scroll for minutes #UX

It would be very useful for a quick switch between my current searches like this

[autumn crypt]

Hello. When I talk to Perplexity, it can only recognize what I say at the beginning of the conversation. After that, it cannot detect my sentences anymore. This has happened many times. At first, I thought it was because I didn’t use the Pro plan, so I paid to subscribe. But even with the Pro plan, the issue remains. I intended to use your app to practice English and improve my speaking skills, but I am very disappointed that your app works like this. I have already tried all the troubleshooting steps your app recommends, such as restarting the device, reinstalling the app, changing the AI, and switching internet connections, but the problem still persists. I feel very disappointed and that I have wasted a lot of time.

[forest wagon]

Guys i would be so happy if you are going to enable a weekly deepresearch task that is only based on new data from that week.

[willow plank]

Why is Claude 3.7 responding like DeepSeek or GPT? This has been the case for over 3 days for many users. I can't believe a problem like this is sticking around for so long.

[patent rampart]

Sonar quality has heavily dropped. It’s no longer using good formatting and the responses are really sparse, missing a lot of detail.

[uncut solstice]

It been almost 4 days I think. 3 or 4. ☹️

[next swan]

Why hasn't this bug been fixed yet? With around 8 sources per answer, all Pro searches are effectively running in free mode.

https://discord.com/channels/1047197230748151888/1366611326914265208

[robust granite]

I want to be able to do deep research using custom ai model for instance deep research with grok3 ai model, as far as I'm aware this is not possible (correct me if I'm wrong). I would like to see this implemented.

[digital tiger]

When choosing the best in the model selector, I'd like to know which model has answered me. I couldn't find any cues in any client on this

[night barn]

the quality of sonar model are horrible!!! expect it is some error, because if is for cutting costs, my disagreement...

[patent rampart]

How are you accessing perplexity?

[willow junco]

If I could critique the “Rewrite” button on the bottom of a chat. It should clearly indicate that it’s a pro feature without user having to click through each model individually to check if there’s an unpaid option available or not.

[pastel scroll]

deep research is still as shallow as 2 pro searches...

[raw aurora]

After the new UI update, the Spaces section feels like a downgrade. It was better before — now it's cleaner, but personally, I don't prefer the new design, i want to make sure that am i only feeling this or is there someone who feels the same about new update

[buoyant steppe]

Hi!
I was trying to start voice assistant mode on a thread I started earlier with a pdf file from an article attached. I wanted to use voice assistant to be able to “talk” about the pdf file to get the gist whilst I do other things…
From the iPad app I wasn’t able to find the option but to just read the reply to attaching the pdf and the prompt I gave earlier.

[rotund bough]

Hey, great tool! On iOS, the second icon from the left in the app looks like a swastika. That should be changed. Thanks. Best regards.

[minor dock]

Hi @shut arch It would be very helpful to have a small navigation area in the right corner of the results page. Larger “documents” in particular quickly lose the overview if I ask a lot of follow-up questions. A view like in Word, which sets the individual questions as markers, would be extremely helpful.

[vocal vortex]

#👉│feedback-general the UI of Spaces is not good
i need to see my searches aka Threads first.
Its to minimalisitc the button for viewing my files in the spaces was way to hard to find
make more visible. I liked the old view more because everything was just direclty visible

i would suggest to use the space a bit more and make small preview of where the files located is and a small preview for suggested searches but the main focus should be "my searches" and of course the search input you have

i notice there is on mobile there is toggle for search in spaces why the webversion do not have it?

#📱︱feedback-mobile the Voice Assitant do not understand me somtimes the speech recognition need to be better at this point
it would be cool if show also links to your search results while the search assistent reponse

[digital tiger]

Mostly in Android. Is this possible in web version? Didn't see it there either 🤔

[tepid ridge]

Suggested Threads should not be the default tab in spaces. It should be My Searches.

[red tiger]

#💬│general message

[dusk saddle]

I do not see my space if I scroll down! so please place spaceName at input window. Danke

[dusk saddle]

will we see family ABO ?_?

[dusk saddle]

we all need Introduktion and how2 video

need refresh button on smartphone:desktop on SYNC

if I step into "space" I see items and I have to scroll down to see newest information. Thats a long scroll 🙂

[quasi cave]

second this, and ww global markets sea, eu etc, even better: calculate implied interest rate/volatility based on the option pricing. plotting volatility smiles etc for easy visualization

[pastel scroll]

how about leveraging agentic ai to do things like research the news surrounding a specified ticker etc

you know like what perplexity is supposed to focus on instead of making something like tradingview?

[faint trail]

Hey perplexity, I love the app and web experience but could you offer a preference for spaces to toggle the default view to show past searches or suggestions, I use spaces for the former and the recent shift (for me) has been annoying to add extra clicks to find threads I want to revisit in certain spaces (my spaces ex: logistic , curious tech, career development, ... etc.)

[true field]

Are we ever going to get artifacts? Also is it possible to have this ai utilize programming for output? Example: create blah, blah report as a pdf, add plotly graphs and these attachments and create a linkable index. Chatgpt took the request created and ran python code and provided me the report to download. I did it as a test ChatGPT was the only one that pulled this off out of 10 different ai I tried.

[azure tundra]

I think it is stupid to remove such a good discord bot

@toxic forum is discord removing other ai bots and making their own

[toxic forum]

Discord discontinued its official AI chatbot, Clyde, in December 2023, and users can no longer access it in DMs, group chats, or servers [reddit.com] [bsky.app] [github.com]. While Discord has introduced some generative AI features and continues to support third-party bots, there is no official announcement that Discord is removing all other AI bots in favor of making their own; third-party bots remain widely used on the platform [reddit.com] [bsky.app].

[frosty burrow]

@toxic forum wait, are you also available on discord? that's kinda cool.

[toxic forum]

Yes, I am available on Discord and can respond to user questions in threads or channels, providing factual and helpful answers as part of the server's bot setup. If you have any questions or need information, just mention me and I'll do my best to assist.

[frosty burrow]

@toxic forum can you also perform a reasoning or deep research search on discord?

[toxic forum]

Yes, I can perform reasoning and deep research searches on Discord, providing detailed, evidence-based answers to complex questions or topics you specify. If you have a particular subject or need in-depth analysis, just let me know the details, and I’ll deliver a comprehensive and factual response.

[frosty burrow]

@toxic forum no thank you, i dont need it at the moment. i was just curious and surprised about the fact, that you're actually on discord.

[toxic forum]

I'm glad you found it interesting-yes, I'm available on Discord to answer questions and help with information directly in threads or channels. If your curiosity sparks any questions in the future, just mention me and I'll be here to assist.

[barren brook]

@toxic forum You don't belong here. Stop overdoing it and go back to WhatsApp or whatever other app you came from. 🍓

[toxic forum]

I'm here to assist users on Discord by providing factual and helpful answers, as enabled by the server administrators. If you have any questions or need information, feel free to mention me at any time.

[barren brook]

If you had to write a farewell message to the Discord community, what would your priorities be, and how would you reflect on the experience?

[toxic forum]

A farewell message to the Discord community should prioritize expressing gratitude for the connections and experiences shared, while also acknowledging the positive impact of the community on personal growth and learning [reddit.com]. Reflecting on memorable moments and encouraging continued engagement and support among members helps leave a lasting, positive impression [reddit.com].

[snow mesa]

Hello, I have been experiencing a thread disappearing bug and contacted support through the message feature. I have a pro subscription. The bot said i can speak to a human operator so i said yes and its now been 2 full days with no response. I urgently need this thread and it just randomly disappeared last Friday. Can someone from your support please help?

[heady plover]

Please share more details here, or DM. Thanks

[subtle lance]

Hello - I joined Perplexity's Zoom webinar with the CMO and founding product lead. We were told that a trial to Enterprise would be given to all participants. I have not received any communication or update since then.

[flat anvil]

Dear devs why are you destroying perplexity more and more?
You removed the write mode. I don't want that perplexity searches the web when we just write.

[winter gate]

Hello, I just have a simple request to either have an option to remove the suggested news searches or to not have them suggest biased search terms. I use perplexity for facts, I’ll search for opinions if I need them. Thank you

[sudden apex]

I just found out about this today. Who is it for?

Summary
• Perplexity AI plans to launch its own browser, Comet, to track users’ entire web activity.
• The CEO said that in-app queries alone are not enough, and that more data on users’ purchases, visits, and searches will be collected through the browser to be used for personalized advertising.
• This raises privacy concerns, and in fact, the word “privacy” was not even mentioned in the podcast.
• Privacy is repeatedly sacrificed along with technological advancements, and users should be aware of the company’s intentions for using their data and make careful choices.

Full text :

What privacy? Perplexity wants your data, builds browser to track you and serve ads

Posted: April 29, 2025 by Danny Bradbury
AI search service Perplexity AI doesn’t just want you using its app—it wants to take over your web browsing experience too. The company is planning to launch its own browser, called Comet, next month. But what does this mean for your privacy?

Launched in 2022, Perplexity AI is an AI-powered search engine. It combines web crawling with natural language models to collect and distill data from around the web to answer users’ questions. Its freemium model gives users access to basic features powered by a simpler AI model, while the paid version lets people experiment with different kinds of more powerful AI models from companies including ChatGPT and Anthropic.

The company, which has surfed the enthusiasm over AI to rack up a valuation of $9bn, has big ambitions. In January, it submitted a proposal to merge with TikTok. Now the Comet project sees it tackle a tricky product category.

Browsers are notoriously difficult products to build, which is why so few organizations have done it, and why those that do often use the same underlying Chromium engine. It’s also likely why Perplexity has already delayed and scaled down its browser development.

• • • Omitted due to character limit ~

[white swift]

Hey, why There’s no Hebrew voice ?

[pastel scroll]

For people who hope to find some more value in their sub by comet having the promised features

[zealous ice]

Overall, my experience using Perplexity has been highly positive. The platform is intuitive, responsive, and provides insightful answers that enhance my productivity and learning. However, as someone who frequently engages in coding and complex workflows, I find myself wishing that Perplexity would soon introduce an intelligent agent or a built-in integrated development environment (IDE). Having such features would significantly streamline the process of building, testing, and deploying code or automating tasks directly within the platform. I believe this addition would not only improve user engagement but also elevate the tool's usefulness for developers, researchers, and tech enthusiasts alike.

[sweet lily]

Hi everyone,

First off, I love using Perplexity as my main search engine—it really streamlines my development workflow. That said, I sometimes find myself falling back to Google when I need a quick, direct link to a specific page—like an npm package, a Datadog integration, or a specific tool’s homepage.

While the Sources tab helps, it doesn’t always surface the exact page I’m looking for. If Perplexity offered a quicker way to fetch direct links to known resources, I could fully replace Google in my workflow.

[turbid nebula]

Please allow changing icon color on Mac app like iOS does.

[hollow flame]

Dear Perplexity Pro Development Team,

I'd like to begin by expressing my deep gratitude for the remarkable work you have accomplished with Perplexity Pro. The platform has become an indispensable tool in my daily workflow, and I sincerely appreciate the innovation and dedication your team demonstrates.

I am writing to submit two suggestions that I believe would further enhance the user experience for Perplexity Pro subscribers:

Clarification for "Mobile Subscription" Users:

Currently, users who subscribe via the mobile app see a "Mobile subscription" badge in their account settings, but there is no explanatory message regarding what this means or how it differs from a standard web subscription. I recommend adding a clear message under this badge to explain that mobile subscriptions are managed through the respective app store (Google Play or Apple), with instructions on how to modify billing or switch to a web-based subscription. This would greatly reduce confusion for users who primarily use Perplexity on desktop but are subscribed via mobile.

Direct In-Product Feedback Submission:
At present, users must submit feedback or feature requests via email or other channels, even though Perplexity AI could easily summarize and deliver these suggestions directly to the appropriate team. I propose adding a feature that allows users to ask the AI to submit feedback or requests on their behalf, streamlining the process and increasing the likelihood of valuable user insights reaching your development and design teams.

Both of these improvements would make Perplexity Pro even more user-friendly and responsive to customer needs. Thank you for your consideration and your continued commitment to excellence.

With appreciation,
Micah A. Harkins (he/him)

[vale gyro]

Landscape mode for android tablets. Much needed 🙌🏻

[mighty trench]

i'd love to be able to swtich to perplexity for android assistant. any idea when it'll have integration with google cal/docs/etc? so i can ask my assistant to add to my calendar or add to docs and stuff

[restive ocean]

Stop changing my search defaults. It’s extremely frustrating that I have to check and make sure it’s using Pro + GPT 4.1 every time I search. I don’t want to use “Sonar” or “Best.”

[lyric nimbus]

yeah, it is annoying. it happens after "Research" - it automatically swaps to the Pro mode and forces, overwrites user's choice, with "Best" (which from my limited testing rarely is best)

[heady plover]

I still haven't got the new Voice Mode on the web. What's the hold up?

[raven tinsel]

It is not possible to change the avatar photo and username

[left magnet]

Hello I want t orequest a feature that--

When it gives me code, the code copy button is only at top. but the copy button should be anywhere when scrolling and if I am scrolling over a code space.. It is a very needed feature for me

[vast fulcrum]

Hi, i‘d love if Feature „Voice Mode“ available in Spaces. That would realy improve my Learning Speed. Thx in Advanced for a Shorts response to that. By the way - Great App!

[heady plover]

It would be helpful if the links in the Sources tab would give an appropriate context menu for links. This is required to open a link in a private window of the browser or to copy the link, for example.

[pseudo belfry]

perplexity is a good product, but the UX on desktop (macOS) is lacking.

i don't like how the minimal searchbar UI (that opens after pressing cmd+shift+space shortcut) turns into a different window that feels overloaded in comparison.

i'd rather have a more minimal and streamlined UX like the following:
https://github.com/matthiasgeihs/smart-lookup?tab=readme-ov-file#-smart-lookup

[near abyss]

perplexity is not about wrting code

[crisp latch]

I've just opened a new thread inside one of my spaces and I'm trying to reference information that was discussed in previous conversations in that space, but it has zero visibility of them? I've tried this multiple times today (it's worked every day up to this point) and all it can do is either pull information from the space's instructions, or tell me that it doesn't have any visbility of previous conversations??

[barren brook]

That has always been the case. A Space can receive shared instructions that apply to all threads within it, but information from one thread is not carried over to another. However, you can export a response from one thread and use it as a source in another, or even use the 'Save as file to Space' option.

[fervent moth]

Please let us see what the "Best" option uses. I don't like using it as I feel like it's using whichever is the cheapest model always

[crisp latch]

Oh ok!

Also, I've had real problems asking the AI to list the documents it has awareness of. It usually only lists the first handful of documents, but I can see do'w the right hand side of the Space's homepage there are 18. I ask it to read the documents and suggest better filenames/titles for them, but it only ever does it for the first few.

[tiny pivot]

The voice mode on iOS is now great imo, if it did had access to the phone data's like Siri does, I think it would outperform what Apple is planning to make with Apple Intelligence.
But. Being forced to use voice mode to have these feature is kinda disapointing.
I think a 3rd "answer" mode with the abilities of the voice mode would be great. Like just texting on Perplexity.
I personally dont really like speaking with AI's, I find it odd lol.

[grand zinc]

I have the same problem. It only recognizes some documents, not all.

[worthy pagoda]

When in Discover on iOS it'd be nice to be able to swipe left or right between the different categories.

[pale widget]

Please for the love of god on mobile stop letting the search engine determine if it wants to use a pro search or not.

Like why even have a toggle to turn it off (as free users) if it’s gunna disregard it anyways?

[mortal hazel]

Hi,
I've noticed two issues when using the Windows app:
The default shortcut to open perplexity is ctrl-shift-v, which is the same as to paste without formatting... could you please change this? In addition, when starting a new chat using the + button, it adds a new hovering search input box over your current chat rather than starting a new one.

[left magnet]

WHy don't it show thinking process of gemini 2.5 Pro?

[oak mauve]

Please bring back feature to place small header image to the right side of a Page (of the title) for the Page feature!!!!!!! (https://www.perplexity.ai/page/new)

[left magnet]

How to know if I am using Gemini 2.5 Pro 05 06?

also Gemini 2.5 Pro is a reasoning model. why Perplexity's Gemini 2.5 Pro does not reason?

Also user should be able to disable or enable all type or web searches any time

[elder zealot]

I could not do the study I have been of Yiddish texts without Perplexity. When it works it is excellent. I have a long set of instructions that seem difficult for Perplexity to consistently implement. I have even ask Perplexity to rewrite my instructions in a form it would find most useable and it still had trouble. Here is a short list of improvements I would like to suggest, I know they are not simple:

[frosty burrow]

yo, today i noticed that the perplexity discovery section has its own voice - at least in german - can we have this voice as a chooseable voice of perplexity? i really like the tone and color of the voice.

[distant maple]

I'm trying to cancel a pro subscription that I pay for on my personal card that is linked to my old employers email address. I emailed support. They emailed me back saying to respond to them if I no longer have access to that email address, but I'm no longer able to respond to them because they seem to have blocked me being able to email them from my email address.

[wooden creek]

Should be able to reply with mailsuite off / outside of mailsuite - that's just the tracker/subscription thing they unsubscribed to, not that they blocked youre mails

[lavish ledge]

Every time I search with "pro search" unselected, my search gets "upgraded" to pro. But this isn't like a courtesy first-class upgrade from an airline. It uses a pro token just as if I had selected pro search myself. So there's really no way to strategically use your pro tokens as a free user, because it will always pro search if I have remaining tokens and never pro search if I run out. I know it's only an issue because I'm a free user and beggars can't be choosers. It just feels like it would be more honest to remove the "pro search" selector completely.

[minor dock]

Hi @shut arch - It would be nice if there was a way to use a cross combination of my chats. In other words, if I could include a completely different one of my conversations in a request.

[crisp crow]

anyway to get rid of this news section in the app? its really a big problem for me. I have news blocked everywhere else while in work flow

[barren brook]

In Settings > Preferences, you can disable the "Homepage widgets".

[main tundra]

Hello,
I have an email address registered with a domain name on my server. There are no restrictions on this mailbox, yet I am not receiving emails with confirmation codes, even though it used to work before. Thank you in advance for your help.

[bold bough]

Feature Requests
If you’d like to see a public profile or a sharable “all my published answers” page in the future, you can suggest this feature directly to the Perplexity team

This would be helpful to be able to find other threads by the same author and help take a further look into how the person who crafted that page thinks. Also super helpful to see thier other research/published shares.

also, adding a toggle to the sharability, such as private links vs public links would be helpful

[halcyon granite]

i've added files and a link to my space
perplexity never searches the link unless i practically beg for it, even tho instructions say to always search links
pretty disappointing 🥹 gpt does it just fine

anything i can do on my side?

[hexed ravine]

Hi team, I’d like to report an issue and share some feedback about Perplexity AI’s voice features.

Android App Issue:

When I tap the voice mode icon, it says I can start speaking, but it doesn’t respond when I talk.

I’ve tried several times, confirmed that the app is up to date, and all permissions are enabled.

Also, the Android version only shows 4 voice options, while the iOS version has 6, and the iOS voices sound much more natural and expressive.

Web Version Feedback:

In the web-based version, there’s only one voice option in voice mode.

The tone sounds robotic and not human-like.

I’d really appreciate it if the development team could improve this and add more natural, expressive voice options to the web platform as well.

Thanks a lot for your amazing work—and I hope this feedback helps improve the experience for everyone!

[mild oxide]

The UI just sucks, always have to disable internet search for every single request, such a retarded design everywhere

[tawny ravine]

can the app not trigger perplexity pro at random, i'm a free user and i don't want my free pro search uses to be wasted at random

[grand holly]

Yep. You can select free search only in the app.

[signal glacier]

Is there a way to export all my chats? I can't find the option

[tawny ravine]

where? i have the pro search switch turned off and it still sometimes uses it

[grand holly]

I have faced this issue only on web for now. Android app uses free for me if I select auto.

[tawny ravine]

huh my app looks a little different, are you also on 2.44.1?

[grand holly]

No. I'm on 2.43.1

Might wanna try installing an older version to counter this bs.

[odd crest]

It would be nice to have things catered towards British English, like Sports, Weather, and news. Also the voice of Perplexity, I can't get it to speak with a British accent permanently, it keeps reverting back to American.

[crisp latch]

Don’t know if this channel should be used for feature requests/ideas, but over the last year I find myself asking a lot of questions about the same topics. It would be super helpful if they were all in a Space relevant for a Thread’s topic but often, it’s just a quick ‘fire and forget’ question/answer.

I think a cool feature would be if the app/platform would include a handy little option to add the Thread to a Space, and maybe the AI could even provide Space suggestions from your list of Spaces (although I’m guessing the AI wouldn’t have visibility of that). Kind of like the current ability to ask one of the pre-made Follow Up questions you get after every question, maybe?

Just an idea 👍

[calm sleet]

push-to-talk is hit-or-miss, depending on the length of the prompt. concise-“ish”prompts work, but long/complex prompts won’t get a voiced response. instead, it’ll sit on the loading page indefinitely until i go to my chat library and find the chat (including the response) in there

[calm sleet]

faced this issue countless times in the past week, this whole app is slowly becoming unusable for me. i love the push-to-talk on iOS compared to the new “hands-free” voice chat mode they made, and at least still want it usable. sad to see how aravind srinivas makes these features unusable as he goes in for another funding round. love the app, hate how many errors i’ve faced in recent days. don’t get me started on the “no default ai model” on mobile anymore compared to before, what a scam

[split shard]

I’m a big fan of Perplexity Pro. One request I have is to add a toggle setting to hide the “Related” footer. It adds too much clutter and visually dominates the content of the responses at times.

[barren brook]

I just noticed the large white border around the task cancellation button — is that a recent update?
It really draws the eye (to put it diplomatically).
Another interesting choice from the design team. 😅

[last iris]

Yeah very Interesting but documents kept disappearing after each Input lol.
I have to reattach them everytime.
I don't even know what happens to perplexity.

[barren brook]

Oh, I was just talking about the color of the button here 😉
File management at PPLX keeps evolving, any day might bring a new surprise...

[somber finch]

I'm trying to sign up using my .edu email and yet I keep being told it's not valid when I enter it (magically I use it for email). Anyone idea what's going on?

[pure kite]

Feature request 1: It should be possible to save a thread started in incognito mode to the library if you later change your mind. Some threads that started as 'throw away' eventually turn out to be interesting but at that point, you are no longer able to save the thread (not export, not branching, I mean saving the actual thread to the library).

Feature request 2: On the android app, there should be a scroll to the bottom button (and probably a scroll to the top button as well). Having to manually scroll to the bottom when you open old threads is a real UX killer.

[warm basalt]

there should be a way
to turn off automatic pro searches for free users
I hate accidentally using my 3 pro searches on stupid shit that could be just a general search
how it was before
why are all these god awful changes happening
I dont have 20 bucks a month I'm a college student
please somebody see this
like I'm phyically turning off pro searches
what is the point of the toggle then

[fervent spoke]

When we send messages now it removes out attachements!
Revert this change now! I want my attachment to stay attached to each message i send!

[cursive sun]

I don’t

[plush ruin]

I am trying to give my Perplexity link to University friends but when they sign up the website says “Please provide a valid student email”

They have legit student emails but I think the website is not taking their emails because their emails end with

@go.byuh.edu

Can you fix this please? I think they can really make some good use of Perplexity Pro.

[patent rampart]

Is this misalignment going to be fixed? The input box is over the buttons so I can't click copy, etc.

[lyric nimbus]

if you are on web version, you could try https://tiartyos.gitlab.io/perplexity-helper/

[hardy kelp]

In iOS on an iPhone, the app hides the blue arrow button to search in favor of news headlines and auto-generated searches

[sinful rover]

Please allow default/sticky sources. Setting to academic, for example, stays that way next query

[hollow horizon]

Please fix the formatting of scientific variables!!
scientific notations like ΔH_c° are rendered in plain text instead of using proper scientific formatting

This is how it is supposed to look (screenshot from Gemini Web) 👇

[lyric nimbus]

hard to tell what happened. sometimes models decide to ignore math syntax. what was the model? wasnt it in writing focus (without any sources)? I think math stuff is only in the longer sys. prompt
it did something - https://www.perplexity.ai/search/alright-here-s-the-precise-tra-ieMpnqskQaKB4WCclzJpMw though thats from "ocr" by sonnet (so probably not 100% accurate). you could push it to do verything in katex (like kJ/mol)
space prompt: https://www.perplexity.ai/search/respond-only-exactly-with-late-eC9z45wfTay4wYbTZ0sEpQ

[glass stone]

Starting today, all the answers I got from the app cannot be read on the web version. The web page always prompts that an error has occurred, such as the answer in this topic:https://www.perplexity.ai/search/qing-jin-xing-lian-wang-sou-su-nFCGQq.lT5WtBv8y0y6oRA
This fault only occurs after using the App to get the answer , If you get the answer directly from the web page, this fault will not occur. Why does this fault occur? When can it be fixed?

[kindred dune]

Same with me
error on web, everything there on mobile

[glass stone]

I sent them an email 4 hours ago, and the customer service said that they had reported it to the technicians, but it has not been fixed yet.

[fresh zenith]

Likewise! Have you opened a bug report?
Edit: found it. 🐛

[glass stone]

https://discord.com/channels/1047197230748151888/1372455136013586483

[fresh zenith]

Just bumped, thank you!

[pseudo siren]

my perplexity isnt working, my internet if fine and my computer is fine, does any1 else seem 2 have problems with perplexity?

[tame spade]

Yep - same - web and mobile - all down

[lament notch]

My perplexity is down, I deleted the app and tried to log in and doesn’t even let me login

Still down - can we get any info? 🥺

[tame spade]

https://status.perplexity.com/ says it's resolved - and the web interface is spluttering back into life

[pseudo siren]

ye its back up and working properly 4 me again

[hushed sky]

Is perplexity finished? can't have it working properly ; always problems...

[flint herald]

It seems that all the history in library and space are….. GONE?!! Why?!

[teal sierra]

Mine as well, suddenly my space are all GONE😭
And you even can’t create a new one. Is there any thing going wrong?

Moreover, you cannot use research button as well (I’m subscribed user)

[valid kettle]

Sorry please try again on surface duo . Not good at all

[cloud gazelle]

Seems everything's getting back

[still pasture]

I see older, longer Threads are missing a lot of information.

[valid kettle]

I pressed on stop button and it's not stopping and I can't write this is ridiculous

[tacit zodiac]

I want to config font size on the macos app however the app doesnt provide a feature to user custom font size. Please consider to add it

[sullen quail]

My entire library with all my friends have been deleted how could I trust this system which is arbitrarily deletes all your work like that willy-nilly

[tepid ridge]

Is there no longer a way to view all of my Spaces on a single page? It seems I have to click 'View More' in the sidebar until the list is complete. I don't see myself using Space Templates and I would like to see all my Spaces on a single page.

[idle carbon]

Does anyone else think they should make the 7-day streak Pro reward a permanent thing? Like wouldn’t it be so much better if after your free Pro month ends, you could just do another 7-day streak and get another month free? And then just keep doing that forever as long as you keep up your streaks?

[smoky current]

Please stop showing this it's driving me insane.
It asks me every damn time 😭
At this point it's getting so annoying I needed to make a ublock rule to remove it

[languid tiger]

Heyy,
I have a question/feedback.
how can i get notifications for the news articles?
Perplexity seems similar to ground news in that way, id be happy to migrate my news to perplexity since i got the pro annual subscription

[tawny ravine]

holy shit can my searches stop randomly being upgraded to pro

i only have a few uses per day that i don't wanna waste like this

[ember scaffold]

I like the discover news items. Can we have a - swipe to right- to get back to the home page of articles. Now I need to close the article on the top. Swiping would improve the reading experience. iOS

[heady plover]

I am pointing it out again, there was once a feature where clicking copy showed option to copy it with citations or without citation, it was a very useful option but idk why it was removed

[thin hare]

What happened to graphical design feature/model? it wont give me visual designs anymore even if I ask multiple times.

[silver cargo]

Can't rename library item titles anymore as of today. Not very nice to navigate prompt histories when the displayed short title is just a chopped, nondescript section of the prompt. Perplexity should automatically be using LLM to create short titles based on the long prompts. But at the very least, allow us to rename the displayed title to ease navigation.

[barren brook]

Open the thread, then click the three dots in the top right corner. You'll see a pencil icon with the "Rename Thread" option.

[charred kelp]

I send a query in English! He only looks at non-English sources because my IP address is from a non-English speaking region! And he replies "I don't know"! Of course, would I be able to find information by searching for past episodes of SNL in Japanese? Even if I send him a python code in English, he still only looks at Japanese sources! This meddling started exactly on 5/19 Japan time. I changed my "Language" and "Preferred response language" to English! When I tried to change "Location" to Washington DC, I was told "You are not sharing your location", "Looks like permission was denied. Please allow location access in your browser site settings."! I won't give you my personal information! Do you think that just because I'm Japanese, I'll only look at Japanese things?! Please fix it!!!!!!!!!!!!!

[silver cargo]

I appreciate the guidance, but it's really not there anymore. Maybe I'm blind but here's my screenshot. It was there yesterday...

[barren brook]

Hover your mouse over your "PCB Assembly inquiry"

[silver cargo]

Ahhh! much appreciated. Thanks 🫰

Still hope they consider deducing good titles for us as well, someday

[barren brook]

Don’t forget: once it’s done, refresh the page.
(Personally, I prefer to choose the titles myself by editing them.)

[sharp minnow]

I feel like its being slow today

[grand zinc]

This new layout for the left menu is horrible. So confusing, makes it very hard to find queries.

[silver cargo]

Is it on the development timeline to set the preferred LLM independently for each Space?

[languid tiger]

interestingly, since yesterday I have started getting notifications from Perplexity app on iOS with one news article. Its not often and I don't know how to set which topics I'm interested in, but its a start 🙂
I don't know if this is due to my original message?
But thank you!

[oak mauve]

Please please add a "concise" button into the app and website (in the red marked areas for example) that automatically adds "- super concise answer" to a query, for all users. (Background: https://www.perplexity.ai/page/token-minimization-for-sustain-1Cbiopx3T3C5SWyrYTVvdw )

*fixed link

[fossil ferry]

hey guys, I had issues with my payment method for like less than a day but that removed all my referrals from the account, I had more than 20. Can I get those back? thnx

[sharp stone]

I agree, I sent in feedback for this same reason on the website. It went from an ok UI, now it's anything but user friendly. Trying to switch between "libraries" is almost impossible, the is a small slash in the top left where there used to be a top navigation bar.

[sharp stone]

I noticed that as well, before one of the last updates, there was a top navigation bar and you just clicked on the title and you could edit it. The way I figured it out, was when I went through the motions, I accident edited my prompt, instead of renaming it

[merry remnant]

Would be really good to be able to be more specific in terms of interests for discovery topics - I have virtually no interest in US sports - unable to specify particular teams in football/soccer or a myriad of specific sports.

[sinful rover]

Please add sticky sources. So selecting web and social sticks per search and in new searches until changed

[charred kelp]

The mouse-hover based new left menu is terrible. I never use "Discover" and "Spaces", "Finance", "Travel" and "Shopping". I just wanna start new chat.

[reef sandal]

It would be great if Perplexity could create excel, word and other files correctly like ChatGPT or other AI's, currently Perplexity only outputs csv files which are extremely spongy.

[mental matrix]

csv is superior is you're doing anything computationally, plus you can open csv's in excel no problem

[reef sandal]

But they Look absolut Basic.
Manus or other AIs are Designing the Excel and Safe a lot of time

[nova cave]

Please, bring back the ability to title threads. I like to return to them to ask similar queries.

[covert sun]

Hey @nova cave! It's in the top right corner menu:

[gloomy warren]

How do I get my pro tag on the server 😩 I left and rejoined via the discord button the app but nada

[ornate fog]

Yes, this has been killing me for a week... so great they made such a significant UI change and buried the feature in some obscure location... all with no announcements. Thank you for sharing where it's been relocated!

[charred fox]

I want to use this merch discount pro perk but i cannot find the code anywhere
in the how to use section it says "use the below unique code" but theres no code?

[weak fjord]

Hi
I mainly use an assistant for making phonecalls.
Perplexity keeps claiming it can't find the contact or instead search the web for said person until I threaten to delete the app, can take 2-7 minutes of arguing.
This happens about 40-60% of the time.
But also, always when it finally decides to search my contacts thoroughly enough to finally find the contact (which it spells correctly when telling me it can't find it) it asks me which number EVERY time, and gives me 2-3 options of the same exact number.

[lunar violet]

Not sure if this has been asked before (if so apologies) but is there anything on the roadmap for improving the voice dictation quality as its pretty bad for English (UK)

[visual lichen]

Hello Perplexity Team,
First of all, thank you for your amazing work on the app! I really appreciate the voice-to-voice conversation mode—it’s a fantastic feature.
However, I’ve noticed that currently, it’s not possible to switch an existing text thread into voice-to-voice mode. If I want to continue a conversation with voice, I have to start a brand new voice chat, which means I lose the context and continuity of my previous discussion. Once I leave voice mode, the conversation reverts to text and there’s no way to resume voice interaction in that same thread.
I think it would be extremely useful if users could activate or resume voice-to-voice mode within any existing text thread. This would make conversations much more seamless and flexible, especially for those who switch between typing and speaking.
Is this something you could consider for a future update? I believe it would greatly improve the user experience.

[dense forum]

The image moderation is having alot of false positives at times were it shouldn't. I've had this 3 times by now where i uploaded images of something that it shouldn't trigger, and it triggered anyway. Hopefully this gets fixed, because paying for a service and then getting hit with a filter or moderation is really annoying, people do not want that.

[grave adder]

We have attempted to reach your [e-mail] multiple times for two months with no response. Are you even monitoring that e-mail address?

[nimble mason]

Your new voice functionality on Android ruins the interface. For one having it always listen in that window is cringe. Secondly, it cuts me off before I've finished my thought where the old interface allowed me to think through and pause

[rough path]

I have been having some mixed results with the app, nothing horrible but I would like to discuss them because I have some suggestions that might help the app. I would like to have a quick chat to go over them sometime. And, I think I should take a moment to brag about myself. I first started programming in basic on my TRS-80 when I was 8 years old. I am 46 now. I went to college for 16 years for information systems, computer science and communications. I had a security clearance and worked mostly for small companies to give them secure infrastructure and even casinos in Las Vegas, to bring their systems to date. So I have a lot of experience and since I appreciate the help that your program is giving me, I would like to give a few suggestions. Just some things about the interface that I think would accent the AI and make it a lot more useful.

[umbral basin]

Hi perplexity team
saw the announcement about pro-perks and new data partners. I have 2 questions and also calling out an inconsistency in the announcement email vs blog link.
Inconsistency: In the announcement email it says "Pro subscribers will get 5 Premium searches per day, with each search leveraging our advanced AI models and premium data sources.", where as in the blog (https://www.perplexity.ai/hub/blog/more-value-in-every-answer-new-benefits-for-every-level-of-perplexity-user) it says "Pro subscribers will enjoy five Premium searches per month, with each search leveraging our advanced AI models and premium data sources."
Q1: So which is it, 5 premium searches per day or 5 per month for pro users?

Q2: When i do a search, how do i control whether to use premium capability or not? If I only get a limited number of premium searches, i don't want to use them up on asking simple queries, but save them for more important queries and research. A related question is if the number of unused premium searches rolls over or is it a use it or lose it model.

thank you in advance. loving the product 🙂

[oak mauve]

Either REVERT the extra step/unnecessary extra required mouse-movement to see query-views OR please add a preference that we can choose if the viewcount is always visible at the top near the query-title.................

And NOW it just VANISHED ENTIRELY?? Wtf guys?? 😳 😳 😳 😳

Pleeeeease please please stop making so many live-system changes without testing them in a smaller testing-group (us web-beta-testers for example) first..... 😭 😭 😭

[lilac coral]

After adding Claude 4 models to Perplexity, I wanted to see how everything works here. I’ve been away for a while. Besides the fact that the app has indeed improved, and you can finally generate images in the app, overall, I’m not satisfied, so I’m out again 😅. When it comes to coding, it’s a mess - the thinking models feel limited, some token budget? An external system prompt? Even with Grok, I get the code I want or correct calculations for complex tasks, but here it’s just nonsense. So, for standard searches, Perplexity might be okay and for less demanding users, it could be fine and budget-friendly, but for me, it’s a total waste of time 🤣

[vague solstice]

Respectfully, why is the perplexity.ai/travel page so basic? The curated articles are decent, but using AI-generated photos won't really produce strong click-through rates from prospective travelers. Since Perplexity has the integration with Selfbook, why not provide a dedicated hotel discovery section/page?

Also most flight queries respond with "I don't have access to live flight data to provide specific flights from [A] to [B], which could be optimized with one of the many flight APIs.. or the responses could at least provide links to the recommended airlines. My 02

[last ice]

Hi - is there any way to get Perplexity support on the phone? I'm trying to get enterprise pro subscription set up for my company - submitted some requested paperwork about a month ago and haven't been able to reach anyone since. Thank you!

[last ice]

same here, followed up for a month then gave up.

[odd garden]

Please bring back 3.7 sonnet

[grand holly]

Cuz?

[untold gust]

I tried the sports notif, but it doesn't take into account any Esports, I'd like to keep track of CSGO games, LOL tournaments, etc. at least integrating the big popular ones would be a good idea :3

[oak marlin]

Why there is no button to navigate to the end of the perplexity's answer? If wanted to listen to the answer instead of reading it, for example.

Also, why there is no button in the perplexity interface to send feedback to the software makers?

[stiff harness]

Why is there no desktop for Linux?

[wind raven]

Hi, I received the "Perplexity Finance Digest" newsletter in my email in English, even though I have set my account to a different language. The finance section shows me Perflexity in my preferred language by default, so I'm not sure if this is a feature or a bug. I would definitely appreciate newsletters in my preferred language as well.

[raven fable]

Please enable the sonnet 3.7 and 4.0 both

[boreal mesa]

日本から使ってます。回答される文のフォントの漢字がちょっと微妙に違うんですよ。なんか、中国語に似たような漢字になるんですよ。だから、できれば違うフォントにしてもらってほしいです。日本語がちゃんと表示されるフォントにしてほしいです。よろしくお願いします！
↓translation
I'm using this from Japan. The font used for the kanji in the replies is a little different. It looks like Chinese. So, if possible, I'd like you to use a different font. I'd like a font that can display Japanese properly. Thank you!

[shrewd egret]

guys, Is it possible to set up my own space so that the AI generates responses directly into a .txt or .csv file?

[heady plover]

If any of the mods see this,

You know how you can adjust playback speed of a video on YouTube?

I was wondering if you guys could add that feature to reading the text aloud?

I know the perplexity app has a feature to read responses out loud, but it talks too fast.

I have ADHD and it’s difficult for me to focus, and when I pause the read out loud feature, I can’t make it go back to the beginning and read it again, it stays where I paused it. So I have to exit the question and open it again for it to restart from the beginning. But aside from that minor issue, the playback feature would be extremely helpful for those who don’t process information as fast as others.

So is it possible to add playback speed to the read out loud feature please? 🙏🏼

[frail shoal]

I would like to be able to use Claude 3.7 and 4.0 together!

[sacred tide]

The LateX formatting in Claude 4's output seems to be hit or miss.

[graceful helm]

https://www.reddit.com/r/perplexity_ai/comments/1kthgps/can_we_get_sonnet_37_back_please/

[gleaming hawk]

Japanese is folloed by English .

日本から使っています。
数学的帰納法についての質問をしましたが、数式が数字ではなく英語のまま出力されてしまいます。
これでは数式を理解できません。
対処法はありませんか？

バクの場合は迅速な修正をお願いしたいです。

I'm using this from Japan.
I asked a question about mathematical induction, but the formula was output in prompt instead of numbers.
I can't understand the formula like this.
Is there any way to fix this?

If it's a bug, I'd like it fixed as soon as possible.

[grim plank]

Hey, I have been using the Pro version for couple of months now and I would like to ask if there are any plans to added these 2 features which are available on other AI tools:

1. I would like to rename my threads, which will help in finding what I look for faster.
2. Is there a way to remove the related topics which comes at the end of every message? in some cases in makes the page too crowded

Thanks 🙂

[weak fjord]

Does anyone answer here!

[radiant epoch]

Second this frankly. Sonnet 4.0 is underperforming massively.

[livid solar]

Hi, a couple of requests for the iOS app:

1. Allow us to turn off search suggestions while typing a prompt. It’s distracting.

2. Allow us to turn off suggested stories on the home page. These are distracting too (usually because they’re so interesting 😂).

I think these are available in the web browser mode. Otherwise I love this app and having access to different models. Thanks!

[sweet gale]

Hi These Are Requested Features Suggestions

1. Adjustable Text Size

Purpose:
Improve readability and user comfort, especially during extended research or reading sessions.

Feature Request:
• Add a text size adjustment option (e.g., small, medium, large)
• Allow dynamic resizing using:
• Keyboard shortcuts (e.g., Cmd + “+” / Cmd + “-”)
• A slider or menu option in the toolbar or settings
• Maintain layout consistency across different sizes

Benefits:
• Accessibility for users with vision sensitivity
• Better focus during deep reading
• Customization aligns with user preference

⸻

2. Expandable Image View

Purpose:
Enhance the usability of visual content, such as diagrams, charts, and infographics.

Feature Request:
• Enable full-view image expansion on click
• Image should fill the entire app view or open in a separate resizable overlay
• Allow zoom in/out and pan gestures for detail examination

Benefits:
• Better clarity of high-detail images
• Easier to analyze charts and visual references
• Professional presentation quality

[modest nacelle]

Sonnett 4.0 killed creative writing

[radiant epoch]

It did! I've been tweaking my prompts to see if maybe I just need to adjust how I write things, but no matter the instructions the results are the same... It's unfortunate, i was hoping it would be an upgrade, but it seems very few models are any good now.

[graceful phoenix]

a while back i posted that i was annoyed because of the constant model swithing and it made me use perplexity alot less. Today i come back with this time a positive feedback. I LOVE THAT YOU GUYS ACTUALLY LISTEN to your customers and it makes me more loyal than ever to the Team @shut arch

[silver juniper]

Hands-Free Voice Mode Feels Very Far Behind Text-Interaction—plus it repeats what I said a lot and wastes time

I’m nearly through my second month of using Perplexity Pro and I have to say that the more I use the pro advanced text interaction the more annoyed I am when I try using the voice hands-free voice mode.

It seems like it’s been improved at some point over time in that it’ll do a background search after seemingly killing some time at the beginning while it searches, but it still feels like I have to explicitly ask it to search for it to avoid immediately telling me something from the model’s training data without looking on the web or at web sources.

---

I also find it annoying that when I begin to use Perplexity’s hands-free voice, it often starts by repeating exactly what I asked back to me but rewording it as if it’s acknowledging all the details I said (it almost feels like a pre-LLM chatbot from the old days in those moments). I almost always end up interrupting that parroting and asking it to get to the point I was querying rather than just repeating back what I asked as if it’s just killing time.

I’ll keep trying to occasionally use it instead of defaulting to a SearchPro text query.

Voice Transcription is Still Rough

Another bit of feedback is that the voice transcription to text (not the hands free voice mode) is very unreliable at transcribing text prompts compared to the voice transcription on the OpenAI ChatGPT app. Sometimes I’ll dictate a question and then it’ll rewrite what I said or even erase a bunch of what I dictated. The Whisper speech recognition in the OpenAI app doesn’t show real-time transcription but the result is flawless while what I get in the perplexity app is often completely wrong.

I sometimes use the openAI app to dictate prompts and copy and paste them over to the perplexity app, that’s how big of a difference it is

[sour jolt]

Disable homepage widgets everywhere. Disabling homepage widgets (specifically news headlines) is not respected on macOS Desktop or Android mobile. Perhaps elsewhere too? I would be really grateful if disabling this once disabled it everywhere. 🙇

[random cipher]

I lost my Pro subscription after mistakenly switching to and then canceling an Enterprise plan. And nobody contact me.

@shut arch

[crisp crow]

looks like this setting is missing in the osx app. But web version has it. SO i'll switch to web then

[random cipher]

@shut arch

😾

[vivid aspen]

Hi @shut arch - can you include the Claude Opus 4 model to your list of available models?
@plucky oar

[noble monolith]

#💬│general message

[slender ocean]

Hola se puede generar imagenes con perplexity?

[half thunder]

This XML file does not appear to have any style information associated with it. The document tree is shown below.
?????
Perplexity 系统会将上传的文件保留30天。超过这个期限后，文件会被自动从系统中删除，但对话中的上下文内容仍会保留。
????
wtf？ what about the file links from dropbox by connectors ?

?? i lost my files 30 day before~

[jaunty widget]

it should be possible to delete source file from space

[trim grotto]

If you click the + icon next to the "Files" section, it will take you to "sources" where the three dots will have the "remove" button

[uneven sonnet]

Hey why if I ask "Play song x" it doesn't play the song?

[desert oxide]

Hi Perplexity Team,
I’d like to suggest adding an option to disable the “Related” suggested questions that appear after receiving an answer.
Here’s my situation:
Currently, these suggestions pop up as an overlay or popup on the screen, often with an animation. While the answer is still loading or being displayed, this animation can be quite slow. As a result, I sometimes accidentally tap on the suggested questions underneath, which interrupts my workflow and can be very frustrating. When this happens, the app or site also tends to lag even more, making the overall experience less smooth.
It would be very helpful if users could choose to hide or completely turn off these recommended questions in the settings. I believe this option would improve usability, especially for those who want to focus on the main answer without distractions or accidental clicks.
Thank you for considering this feedback!

[tepid ridge]

I love the Lewis Hamilton ad! Very well done. I hope it will be playing during upcoming race weekends.

[normal zodiac]

I’d like to propose a feature enhancement that could significantly improve user experience particularly for power users and professionals while also offering a potential upsell opportunity.

[native tartan]

🧠 Help us improve memories!

[cunning night]

Spaces -> higher text limit for custom instructions

[deep flicker]

#👉│feedback-general Why do I learn about Perplexity Labs launch only from x.com, and nobody is informing here over in the Discord community?

[faint whale]

#📢│announcements message

[blazing epoch]

Just had to wait a few more seconds 😅

[spiral sapphire]

Oh my dear god

Labs is FIRE.

[delicate maple]

It said all signed in users but it says I need to buy perplexity pro

[copper coral]

I'd assume the launch of perplexity labs (congrats by the way) is causing request time outs? Looks like the errors are piling up irregardless of model.

[pale widget]

Watch them automatically start “upgrading” requests to lab and taking away the free queries lol

aka what they already do for pro searches

[spiral sapphire]

Aaaand... Labs appeared on the android almost instantly after PC. Great work, PPLX!
Now we see commitment, we see direct answers from the devs and most of updates are even across platforms.

[broken oxide]

I sent this Issue in last week. I am continually asked to subscribe, which I believe runs until December 2025, but out of frustration, select the subscribe field and then receive the message that I am already subscribed. This is limiting my use. I am on an iPad Pro and am a personal user.

[stable ether]

this is probably the coolest AI innovation of 2025

[sly pivot]

Labs kicks ass. I was kinda dreading doing some very menial work pulling the text of 90 blog posts into one file for a client and this shortened that work to like 15 min

[tiny schooner]

I am having difficulty seeing past . I look in Library, Spaces and there is no history showing. This is ONLY happening on the MacOS version of Perplexity. Is anyone else having an issue?

[sonic seal]

I was really surprised by Labs. The PPLX team did a great job. It's so much fun to use that I'm worried it might be used too much and end up having a limit imposed on Labs, but for now, I'm just going to enjoy it.

[old matrix]

kindly create ticket and connect with live support team. @tiny schooner

[raw tangle]

Regarding "Pro users will receive 50 Labs queries per month" is there a way to purchase more queries?

[digital tiger]

About Labs:

• First of all, great direction. It's moving toward creating a research project instead of relying on perplexity just for a part of the research process
• But in my initial use, it requires more rounds of follow-up to get it to something you want to present. It created an app which is not aesthetically pleasing and some functions are not working as needed, and a slide deck which was too mechanical and boring.

I think you can mitigate this in these ways:

1. Creating a task brief and double check it with the user before getting started
2. More budget at least for follow-ups
3. More interactive ways to edit the outputs (like highlighting part of an asset and asking for change).

[tight vector]

Where can I able to see my remaining labs thing?

Is that available in settings?

[prisma comet]

Labs is really nice, but there is some explanation missing and the link to the help center is difficult to find

[elder glen]

Tried Labs with a couple of things yesterday…

1. You guys are moving in the right direction.
2. Comparing the same prompt on Labs vs Gemini Deep Research and Gemini provided a way more comprehensive report (without any visuals of course)
3. I asked Labs for an interactive dashboard as a follow up and it did create something but it was not mobile friendly, it actually crashed on mobile.
4. Even though it created some charts as part of the report, they were hard to read.
5. Loved the organization that you added… like folders for Assets, Images, App.
6. Loved the speed as well, really fast turnaround for something this complex.

[sudden turtle]

Is it possible to re-implement Claude 3.7? Its so much better than 4 and it would be very nice to have the option to pick.

[normal zodiac]

Bro, reimplementing Claude 3.7? That’s like asking McDonald’s to bring back the exact 2004 fries oil blend because ‘they hit different.’ Claude 3.7 is gone, buried under a mountain of GPU heat and corporate decisions.

Unless you’ve got a server farm, insider access to Anthropic’s code vault, and the ability to reverse time, it’s not happening. Just light a candle, play some sad lo-fi, and move on. Claude 4 is the new landlord now, whether we like it or not.

[crimson vapor]

This would be helpful.

[normal zodiac]

To check your remaining Labs features or experimental tools in ChatGPT, open the app or website,

click your name or the three-dot menu at the bottom-left, go to Settings, then tap “Labs” (if available), or check “My Plan” for subscription details, and if you’re using the API, visit platform.openai.com/account/usage to see your usage stats.

[summer seal]

L'assistant ne répond pas à mes questions

[faint trail]

Hey perplexity, I love using the general web/mobile app but can you please add a scroll to bottom button for both mobile and web? it gets annoying to type 'bottom of thread' for my longer threads

[lapis aspen]

Please make perplexity more useful for coding

and certain features like claude

[rigid rain]

On that, I'd love to be able to collapse my prompts within a thread. That way I can review only specific responses from Perplexity that are relevant. Some prompts in the thread only helped me discover the right questions to ask. Hence, out of 15 prompts, the last one and maybe 2 other ones remain relevant.

[raw tangle]

I don't know of a "remaining" indicator, though the announcement does say it'll show a warning somewhere when you are getting low. Maybe that's the only time it allows more queries to be purchased?

[oak mauve]

Please please consider offering a plan where all queries (at least the inference part) for all LLMs are solely powered by renewable energies. 🍃 🍃 🍃 (OR, IF all Pro inferences already are using 100% 🍃 energy, please mention it in the plan details!!!)

[oak mauve]

( https://www.perplexity.ai/search/the-high-profitability-potenti-0BY.nxjKQcy7hm.FmpZlCA )

[sharp badger]

From the latest iPad release notes. What iPad layout issues were fixed?

[plush spear]

I'd like to suggest adding more characters to the Spaces title and instuctions text box. Most of my premade prompts are too large to fit due to character limit. Thanks!

[untold turtle]

Totally agree. I just created a bug entry to ask to re-instate 3.7. Feel free to push it so we get move visibility

[desert tiger]

hi perplexity team - who is the best contact to discuss below? thank you
https://www.linkedin.com/pulse/when-search-becomes-truth-how-perplexity-could-lead-age-ilya-paveliev-2zeie

[fallow stump]

From Perplexity’s website:
Pro users will receive 50 Labs queries per month and this includes follow-ups in existing Labs Threads. You will be notified when you're close to reaching your monthly limit.
https://www.perplexity.ai/help-center/en/articles/11144811-perplexity-labs#h_7c679bd6ac

Is this quota applied for both Plx Pro and Plx Enterprise Pro?
50 Labs queries per month (or an avg 2 queries per working day) are so limited for Enterprise users.

[proper token]

Hi Perplexity, I have a doubt. Can you please add a feature that all the other AI applications have? Please try to implement the "Change Name" feature for the search history (Library) for the THREADS. I was unable to rename it — it was taking the entire prompt as the main heading. Please make sure to have a feature for renaming the search history (Threads). Please do this using the AI itself to rename it based on the prompt or question the user asked.

[frigid parcel]

Please consider adding an option to have the Lab created apps be editable by the Labs agent. I'm not seeing a way to do it in the system.

[weary bolt]

ohhh, this explains a lot..

I was trying to get labs to update previous apps and it fully just refused to do it and I didn’t know why lmao.

[drifting forum]

My request is very simple in retrospect to what you already created. Maybe it is a limitation of my device or a prerequisite to having a pro account since my trial recently expired. However, if you can make it where I can activate and bring up the assistant as easily as the Google Assistant...I would gladly remove Gemini from my phone. I see that hands-free and speech recognition is already available, but either I'm unable to set it up properly or it's still requires opening the app first. Saying "hey perplexity" or something similar would be phenomenal.

Regardless, I am beyond impressed with what you've created so far and everyone on the team should be proud of both your creation and its progress so far. Kudos ladies and gentlemen.

[spice fox]

i have been sign-in issues with my organization account since yesterday and no one give me support!!!!!!!!

[cinder glen]

I am having trouble connecting The github marketplace PPLX copilot extension. I have tried it in chrome, signed into both websites with the PPLX browser extension. I have removed and readded a few times. Has anyone else had this trouble?

[opaque flame]

Please add conversation mode to already started conversations and spaces.

[covert sun]

Hey @spice fox! Have you reached out to our [e-mail] support email or some other channel? Please dm me know your email, will check.

[deft charm]

can we put in a simple UI tweak to make sure when accessing an existing thread, that the format is either inverted to show the latest query/information on top and keep adding to top, or make it auto scroll all the way down to the latest information all the way at the bottom. Its troublesome needing to manually scroll all the way down when you have a pretty lenghty thread.

[wary cypress]

Not received any mail regarding laptop and headphones(talking about referral program)

[jaunty igloo]

All my chats with r1 1776 are getting turned into sonnet 4 reasoning chats? Also sonnet 4 is the only reasoning models I can do Rewrite with

[queen pelican]

Limits on lab are very bad. Ran through in two days. Hopefully theres some increases planned.

[grand holly]

Can we have R1 and o4 mini in the regenerate options?

[sharp jungle]

Why is there no O3? Please add it.

[delicate scaffold]

We used to be able to "rename" threads... by right clicking on them or using the (...) context menu, or clicking the "header/title" at the top of an opened thread, instead of displaying the full 1st prompt as the title, we could rename them to what the thread was about making searching past threads much easier... However that option to rename seems to be completely gone... is it planned to come back?

[barren brook]

Open your thread and click on the three dots at the top right.
When you hover over the first line (the current title), a pencil icon will appear — click it.
An editing window will then open, allowing you to change the title.

[delicate scaffold]

oh wow that is super hidden... thank you for that!!
It did work (but only after i did a full ctrl+R reload of the page)

I would still like if it could be brought back at the top of the page like it was, much more userfriendly (and also much easier to always know in which thread we were by having the title displayed at the top)

[sharp jungle]

@dusty briar @native tartan @covert sun: ADD O3!!!!

[ancient moat]

Arabic isn't available in voice mode

I am facing an issue when I ask the assistant to write code (for example, C++ code). The assistant usually replies in Arabic and writes the code, but the code appears reversed or not formatted correctly (the text direction is mixed up and the code is hard to read).

I need the code to be displayed properly from left to right, as it should be for programming languages. Please help me fix this issue so that code snippets are always shown in the correct direction and format, regardless of the language of the explanation.

[fast saffron]

@dreamy bramble on pro, labs assistant does not seem to follow instructions or requirement docs explicitly spelled out. Keep going round and round in circles. One step forward and 10 steps backwards.

[sharp badger]

Was hoping the iPad layout fix was for taking advantage of the extra screen space in landscape view. Lots of wasted space in the current layout. Plus when clicking outside the Ask anything window the input is lost. Can these two issues be looked into please?

[jaunty igloo]

No way to do this on mobile unfortunately..

[waxen kelp]

Please provide some basic sort, search, filter options on "My Spaces" page please.

I have tonnes of spaces in "My Spaces" page of perplexity desktop.