thorn mauve
@everyone Welcome to the new LMArena server! 👋 Since our graduation from LMSys, we're now moving all Arena projects, and leaderboard updates here to better serve the community. Expect more server updates and community surprises to roll out over time. ✨
All Arena related channels in the LMSys server, will be deprecated starting on March 14, 2025.
Learn more about our graduation here: https://x.com/lmarena_ai/status/1842982750095278482
X
jaunty rain
🚨 New UI Design Access - Heads Up! 🚨
@everyone It looks like the cat’s out of the bag—our new UI design access has been shared! 🐱💨
Here are a few things to keep in mind:
🔸 Super Alpha Stage – Expect breakage, data loss, and plenty of bugs along the way.
🔸 Constant Changes – Things may shift at any moment as we continue refining the experience.
Thanks for bearing with us—your feedback is invaluable! 🛠️ 💙
If you're still interested, then check it out here:
https://alpha.lmarena.ai
pw: super-alpha
PLEASE Give us feedback here: https://forms.gle/8cngRN1Jw4AmCHDn7
and 🪲 report bugs here: https://airtable.com/appK9qvchEdD9OPC7/pagxcQmbyJgyNgzPx/form
Please note: This super alpha is currently optimized for desktop only! Mobile is not yet ready at this time. 🖥️
thorn mauve
🙏 Bug Reports & Feedback on the new Desktop UI – Help Us Improve! 🛠️
@everyone
New UI here: https://alpha.lmarena.ai/
pw: super-alpha
Whenever possible, please submit bugs and feedback using the links below so we can prioritize and triage quickly.
💠 Feedback : https://forms.gle/8cngRN1Jw4AmCHDn7
🪲 Bugs: https://airtable.com/appK9qvchEdD9OPC7/pagxcQmbyJgyNgzPx/form
💬 Want to discuss the new UI? Join the conversation in #new-ui-feedback feedback!
Your insights are greatly appreciated—thanks for helping us make things better! 🙌
thorn mauve
@everyone Our First Community event! 🎉
Check out the Events tab 🗓️ in this server for more information on our first Stage event on Thursday. We’re excited to walkthrough and chat about the new Desktop UI with you all!
Don’t forget to submit and vote for questions in advance for the Q&A. Hope to see you there! 👋🏽
thorn mauve
@everyone 📢 If you haven't already, let us know if you're coming to our first event tomorrow! Details above! 👆
BTW, we heard you in #new-ui-feedback and will demo an update that just rolled out around the voting UI. Check out the preview here and go test it out in the alpha! ✅
➡️ Reminder to submit and vote for Q&A here: https://app.sli.do/event/7bThoD3UhfcLdJTLiteEgz pw: super-alpha
thorn mauve
@everyone 📢 Today is the day for our first community event!
🎉 Head to The Arena Stage channel for a live walkthrough on the new Desktop UI, and a little Q&A. The event will start at <t:1741894260:R> and be about 30 mins long.
We can’t wait to see you there! 😃
thorn mauve
@everyone Thank you to everyone who joined our first Discord community event!
We want to make these even better, and your feedback is key. Tell us what you loved, what could be improved, and what you’d like to see next! Didn’t attend? You can still let us know what you would like to see:
💡 Why share your thoughts?
💠 More events tailored to what you want
💠 Better content, speakers, and formats based on your input
💠 A chance to shape the future of our community
👉 Take a minute to fill out our feedback form here: https://forms.gle/Hr9xgSTWnyLVR9tC8
Your input makes all the difference—let’s make the next one even bigger and better! 🚀
thorn mauve
@everyone New Desktop UI – still in Alpha!
Your feedback so far has been incredibly helpful!! 🙏 🙌
Please keep testing the Alpha here: https://alpha.lmarena.ai
🔑 We've removed the password to make it easier for you!
REMINDER: as an Alpha UI it's got limited features, but frequent updates. Be sure to go to the current site for the freshest leaderboard data, features, and the full set of models.
Please share feedback and bugs through these channels as it makes it easier for us to reproduce errors and prioritize requests, we're hard at work already implementing many of them! 🚧
💠 Feedback : https://forms.gle/8cngRN1Jw4AmCHDn7
🪲 Bugs: https://airtable.com/appK9qvchEdD9OPC7/pagxcQmbyJgyNgzPx/form
💬 Want to discuss the new UI? Join the conversation in #new-ui-feedback!
thorn mauve
@everyone We love all the great feedback on the Alpha Desktop UI—keep it coming! Your insights are shaping the future of Arena, and we truly appreciate it.
As we continue working on some major improvements to bring everything up to speed with the current site, we’re re-implementing the password to keep access limited to you, our community. This is still an early-stage build, and not yet intended for public use.
🔑 New password: still-alpha
Thanks for being part of this journey with us! Keep testing, sharing feedback, and helping us make Arena even better.
thorn mauve
Hi @everyone!
Thanks so much for continuing to test the alpha — we’ve been busy making updates based on your feedback. Here’s what’s new in the alpha:
🔗 Please keep testing the Alpha: https://alpha.lmarena.ai/
🔑 Password: still-alpha
Your feedback is super valuable — it helps us reproduce bugs and prioritize your requests. We’re already hard at work on more improvements! 🚧
💠 Feedback: https://forms.gle/8cngRN1Jw4AmCHDn7
🪲 Bugs: https://airtable.com/appK9qvchEdD9OPC7/pagxcQmbyJgyNgzPx/form
thorn mauve
📣 Alpha Arena Updates @everyone
Try it out here https://alpha.lmarena.ai/ (password: still-alpha)
Thank you for continuing to test! Keep the feedback and bugs coming 🙌🏽
💠 Feedback: https://forms.gle/8cngRN1Jw4AmCHDn7
🪲 Bugs: https://airtable.com/appK9qvchEdD9OPC7/pagxcQmbyJgyNgzPx/form
thorn mauve
📱Alpha is now mobile-ready! @everyone
You can now test the new Arena Alpha UI right from your phone. Whether you're on the go or just prefer mobile, the experience is now optimized and ready for you.
We know a lot of you have been waiting for this — so now’s your chance to dive in and put it to the test!
🔗 https://alpha.lmarena.ai
🔑 Password: still-alpha
As always, your feedback helps us improve fast:
💠 Feedback: https://forms.gle/8cngRN1Jw4AmCHDn7
🪲 Bugs: https://airtable.com/appK9qvchEdD9OPC7/pagxcQmbyJgyNgzPx/form
Come chat in #new-ui-feedback and let us know what you think!
thorn mauve
@everyone Thank YOU for all the rich feedback on Alpha UI in the past week!!
Please keep testing, and sending notes!
🔓 No password needed anymore at https://alpha.lmarena.ai/
Reminder this is still an early version, so some features are limited, but updates are coming fast for Desktop & Mobile.
For the latest models and leaderboard data, use the main site.
Got thoughts or found a bug? Let us know here:
💠 Feedback: https://forms.gle/8cngRN1Jw4AmCHDn7
🪲 Bugs: https://airtable.com/appK9qvchEdD9OPC7/pagxcQmbyJgyNgzPx/form
💬 Chat about it in: #new-ui-feedback
thorn mauve
📣 Big News: We are starting a company + New Beta is Live! 💥
Hey @everyone — we’ve got two big updates to share today:
1️⃣ We are starting a company to support LMArena!
We began as a scrappy academic project out of UC Berkeley, and thanks to all of YOU, we’re taking the next step that will allow us to stay committed to improving the platform you’ve helped us build. LMArena will stay neutral, open, and accessible to everyone. Read more here: https://blog.lmarena.ai/blog/2025/new-beta/
2️⃣ Beta is LIVE!
We’ve been listening closely to your feedback from the Alpha, and today, we’re releasing a Beta version of the new LMArena site:
🔗 https://beta.lmarena.ai
📝 A note for those that were testing the Alpha: saved chats won’t carry over in this version. We know that’s a bummer, we’re still fine tuning this - thanks for your patience as we continue to improve! 🤗
☑️ Since Beta will still be a bit buggy, we will be testing the signal quality. Votes are currently being stored, and we'll start to include them properly as the signal quality increases. Your feedback helps us make the evaluation stronger, so please keep voting!
🔗 Try the Beta: https://beta.lmarena.ai
💠 Feedback: https://forms.gle/8cngRN1Jw4AmCHDn7
🪲 Bugs: https://airtable.com/appK9qvchEdD9OPC7/pagxcQmbyJgyNgzPx/form
💬 Chat about it in: #new-ui-feedback
Thank you for being part of this. Let’s make it great together! 🤝
thorn mauve
The feedback is coming in hot for Beta! 🔥 THANK YOU @everyone 🙌🏽
...and we are responding!!
We now have:
🌓 Dark/Light mode toggle in the top right
✂️ Copy/paste images directly into the prompt box
✨ A few polish items in the leaderboard
Keep sending the feedback! 🙏🏽
thorn mauve
Hi @everyone, do you have time for a quick poll? 👋
Every vote helps us understand the community better - thanks as always for your help! 🙏🏽
#general message
thorn mauve
@everyone Around this time 2 years ago, this community helped us launch our very first Arena leaderboard!
Today we’re celebrating what you all have built together on LMArena! 🥳 👏 A few fun stats:
☑️ 3M+ votes!
🤖 400+ models on the leaderboards!
📊 300+ pre-release evaluations!
📝 10+ open datasets for prompts and user preferences
Take a look at our OG leaderboard below - back when we were still developed Vicuna🦙
Read more about this community’s impact in our blog post: https://blog.lmarena.ai/blog/2025/two-year-celebration/
tweet: https://x.com/lmarena_ai/status/1916620122342695363
minor tide
Hi @everyone! 
Building community is paramount to LMArena's mission. That's why our team is investing more time and energy into creating a space for those interested in making an impact in AI. I'm excited to share that I'll be stepping in as this Discord's community manager! You'll be hearing more from me as we implement improvements to help grow, engage, and protect this space - all in service of building our AI community.
Building community requires... well... community! So I'd love to hear your thoughts through this survey about possible changes you'd like to see happen in this Discord.
It's nice to meet you!
minor tide
Hey everyone Quick heads up that over the next few days, we'll be making some server changes focused on new member onboarding, channel structure, and mod reporting. Listening to community feedback is incredibly important to us, so please let us know how you’re feeling about these changes. Don’t hesitate to reach out with any questions!
Also, many of you have requested independent scrolling, here's a sneak preview of what we're working on!
minor tide
@everyone As some of you may have noticed we've implemented a few changes to the server in an effort to make a more engaging and protected space. It's important to note that we're driven by community feedback so if you have thoughts on these changes fill out this form!
Server Structure
- We've added a new Forum Category that's intended for gathering feedback, troubleshooting issues, and model requests. This is intended to replace the #new-ui-feedback and #arena-feedback channels to better organize and track issues, feedback, and requests! Note if you're not seeing these new channels you may have to enable them in:
Channels & Roles -> Browse Channels.- New Roles have been created! In the
Channels & Rolessection you'll find a few questions that'll auto-assign new roles. These roles will allow for more targeted announcements ensuring you all are getting the information that's most important to you!- The Server Guide is now where you'll find our #rules & #information-desk channels.
Server Guideis located at the top of the channel list.- Channels that weren't getting much use are being moved to the Archived category. We're looking to create a more engaging server and phasing out channels with little use will reduce clutter.
Moderation
- For immediate needs pinging <@&1349916362595635286> role is available.
- For issues you'd like to flag privately you can now **send a Direct Message to the ModMail bot **which you'll find at the top of the
Member List. This will provide more options for members to report bad behavior.- We've updated our #rules a bit so be sure to give that a peek. These changes are intended to keep discussion more on-topic and create a more inclusive space.
🪴 The Future- Our plans are to host events on a more regular basis! Staff AMAs, contests, and casual game/activity nights are all things to look forward to. So keep an eye out!
minor tide
<@&1372208635530448926> new models are now live on the beta site! Go check em out! 
- mistral-medium-2505
- claude-3-7-sonnet-20250219-thinking-32k
- amazon.nova-pro-v1:0
- command-a-03-2025
minor tide
<@&1372208524230397962> Since the debut of Mistral Medium 3 we've seen some impressive moves on the leaderboards:
- #11 overall in chat (a +90 leap from Mistral Large)
- Top-tier in technical domains (#5 in Math, #7 in Hard Prompts & Coding)
- #9 in WebDev Arena
minor tide
@everyone Sharing some news about the future of LMArena! 
Next week, we are planning to flip the switch and change the current site to the beta site! Additionally, we’re excited to share the news that we have received $100M in seed funding! What this means is hiring more people, increasing site performance and incorporating community feedback faster into the experience. Community feedback will continue to play a crucial role in helping shape the platform.
You may also notice the fresh coat of paint on the server! Along with the new look we’ve added some custom emojis - 
We’re planning to host a staff AMA soon! If you have questions for the staff please fill out this form. We’ll be sure to announce the date/time when confirmed. We also plan to record the event for those that are unable to attend it live.
Quick note to use the new #1372230675914031105 and #1343291835845578853 forum channels for any feedback or bugs you’d like to share with the team. Also, be sure to grab those server roles in the Channels & Roles section if you haven’t already!
minor tide
<@&1372208635530448926> happy to share that **the next generation of Claude is in the Arena! **
- Claude Opus 4
- Claude Sonnet 4
minor tide
@everyone Happy to share that we'll be hosting our first Staff AMA at <t:1749229200:F> ! **LMArena's Cofounder & CEO Anastasios Angelopoulos ** will be first up in our series of Staff AMAs we plan to bring to this community on a more regular basis. For those unable to attend live, we plan to record the event.
minor tide
<@&1372208524230397962> <@&1372208635530448926> Last summer, Style Control was introduced to disentangle model response quality from stylistic factors like length and markdown formatting, helping to better reflect core capabilities. Many community members agreed this provided a clearer assessment, so as of today, Style Control is now the default view. You can still disable Style Control through the filter options.
<@&1372208590248742964> be sure to check out our research into how style influences votes here!
minor tide
<@&1372208635530448926> The highly requested independent scrolling feature is now here! Each response area is now individually scrollable. Thank you for all your feedback and enjoy!
minor tide
<@&1372208635530448926> New a16z podcast here! Our cofounders sat down with the lead investor at a16z to talk about LMArena and the future of AI evaluation. The team discusses how LMArena evolved overtime, why subjective data is crucial, and what it means to build a CI/CD pipeline for large models. Watch the episode here.
We'll be hosting a watch party <t:1748619000:F> in #1340554757827461215 for anyone that wants to join!
minor tide
<@&1372208635530448926> DeepSeek R1-0528 is now in the Arena! Go check it out.
minor tide
@everyone We want to see what you all can make with LMArena! Happy to share that today we’re announcing our first AI Generation Contest. Each month we’ll be crowning a new <@&1378032433873555578> through community voting.
How will it work?
- You may have noticed the new #june-contest channel which is where you all will “submit” your creations and discuss what others submit
- To submit simply post a screenshot of what you’ve created through LMArena to #june-contest
- On June 20th submissions will be closed and we’ll circulate a survey allowing you all to vote on which ones you like best
- The person with the highest score will be declared our winner!
- 1 month of Discord Nitro
- <@&1378032433873555578> role! You’ll hold onto this exclusive role (for at least a month) which will be hoisted towards the top of the member’s list showing the
worldserver your accomplishment!
- Submissions must be done through Battle Mode, submissions created through side-by-side or direct chat will not be accepted
- Your submission must include both the left and right response
- Your submission must be after you’ve voted for which response you prefer meaning the models should be revealed in your submission
- Only one submission per person
example provided below
Let’s get cozy! Warm beverages, fluffy blankets, and overall snug vibes is what we’re looking for - but at a desk. Get creative and show us what you think would make a cozy environment. This month’s contest will be for image creations only.
Happy submitting and good luck!
minor tide
<@&1372208635530448926> Our leaderboards were recently updated. Go check them out! https://lmarena.ai/leaderboard
A reminder that this Friday we'll be hosting our Staff AMA. We are planning to record the session for those that can't make it live. https://discord.gg/XkfsbYWX?event=1375223423009165435
minor tide
@everyone We're planning to take the feedback process to a whole a new level - and you're invited to apply!
The LMArena Test Garden is going to be our new private feedback program that'll invite selected users to get exclusive sneak peeks at features, design mocks, and ideas the team is considering implementing for early feedback to ensure we're on the right path. For those who are exceptional at providing feedback, and would like to see what we're cooking up, this program is for you!
If selected we'll followup privately for next steps.
minor tide
@everyone
We have a contest running right now! Don't miss out on the chance to win! Post your creations to the #june-contest channel. More details here - #announcements message
Excited to provide feedback and interested to see the things we're building behind the scenes? Apply to our Test Garden! Apply here!
Thank you all who attended last week's Staff AMA! We look forward to bringing more of these in the future. Please share your feedback here!
minor tide
The team is aware of a widespread issue where models aren't providing a response but instead are erroring out. We are working on a speedy fix. Our apologies for any inconvenience this causes. I'll update this message when it's fixed.
It's fixed! Should be working again. 
Don't hesitate to @ me if you're still having issues.
minor tide
@everyone Help us determine June's <@&1378032433873555578> for Cozy Desk by voting here!
Reminder for what we're looking for:
Let’s get cozy! Warm beverages, fluffy blankets, and overall snug vibes is what we’re looking for - but at a desk. Get creative and show us what you think would make a cozy environment.
minor tide
@everyone <@&1372208635530448926> <@&1372208524230397962> A new leaderboard is now live! Driven by community votes we’re happy to share that the Image Edit Leaderboard is now available with 7 models currently filling the ranks! With the Image Edit, you can upload an image and directly compare each model’s editing capabilities.
Go check out the Image Edit Leaderboard here!
minor tide
<@&1372208635530448926> New models added to LMArena!
- mistral-small-2506
- imagen-4.0-ultra-generate-preview-06-06
- ideogram-v3-quality
minor tide
<@&1372208635530448926> New model added to LMArena!
- grok-3-mini-high
minor tide
<@&1385704581677191278> To celebrate the launch of the Image Edit Leaderboard let's incorporate some Image Edit capabilities into our July contest!
How does it work and what are the rules?
- Submit your entry by sharing a screenshot of what you've created in the #july-contest channel.
- On July 25th submissions will be closed and we'll circulate a way for the community to vote for our winner.
- Submissions must be done through Battle Mode & use the Image Edit functionality. You must use both an image and text to inspire something new.
- Your submission must include both the left and right response.
- Your submission must be after you've voted for which response you prefer meaning the models should be revealed in your submission.
- Example here.
What could I win?
- 1 Month of Discord Nitro
- Become the newest member to receive the <@&1378032433873555578> role!
July's Contest Theme - Out of Place Objects in Space! 🪐 🚀 ☄️ Sci-fi space environment is a must, but include something that clearly doesn't belong. Confuse us!
Help me send a big congrats to @tawny basalt for being our very first <@&1378032433873555578> !! A very cozy desk was achieved, check it out here!
minor tide
<@&1372208635530448926> New model added to LMArena!
- seedream-3 (text-to-image)
minor tide
minor tide
@everyone <@&1372208635530448926> We're excited to share with you all that we've made some light UI improvements that are now live! Our intention with these improvements is to make the overall experience more polished, intuitive, and delightful. Many of these improvements were inspired by community feedback.
What’s new?
- A more streamlined interface. The core chat UI is now a more focused and minimal experience that reduces visual clutter.
- A more compact sidebar. This provides quick access to About Us, How It Works, Feedback, Leaderboards, and other sections.
- Leaderboard tab now has improved navigation. Accessing the leaderboards you care most about can now be done faster.
As always, community feedback is crucial. We'd love to hear what you think about these changes here.
We have an exciting July ahead with numerous changes and improvements in the pipeline that we're eager to share with you!
minor tide
@everyone <@&1372208635530448926> A new modality has been added to LMArena. Check out Search Arena here!
7 models with search capabilities are ready and waiting for your testing. Note to have the Search modality selected in the chat box first.
- Grok 4
- Claude Opus 4
- Sonar Pro High & Reasoning Pro High
- o3
- GPT 4o-Search Preview
- Gemini 2.5 Pro Grounding
Learn more about what Search Arena has taught us about human-AI interactions on our blog post.
minor tide
Psst… we’ve got a surprise for you.
An experimental Video Arena bot is now live, and you can try it right here only in this server!
Generate videos and images with top AI video models with the LMArena bot:
🗳️ Vote on each other’s creations
🧠 Learn how to use it in #1397655624103493813
💬 Share feedback in #bot-feedback
minor tide
<@&1372208635530448926> New models added to LMArena!
- GLM-4.5
- GLM-4.5 Air
minor tide
@everyone <@&1372208635530448926> <@&1398740297521037332> We’re launching an experimental Video Arena here on Discord. Generate videos with the top AI models for free, and compare their results right here in our community server.
Learn how to use this bot in #1397655624103493813 and start generating in #video-arena-1 #video-arena-2 #video-arena-3.
What does the bot do?
- With the LMArena bot you can generate videos, images, and image-to-videos via the bot. Similar to battle mode you’ll given two generations and anyone will be able to vote on which they prefer. After a certain number of votes, the bot will reveal the models.
Why is this being considered an experiment?
- There are a lot of firsts with this one. First time access is exclusive to our Discord server. First time others are able to vote on other’s generations. First time you all will be inspired and chat about what you and others generate. That being the case, we're excited to hear what the community thinks!
To celebrate this new milestone, join us for a Staff AMA with the Bot’s Developer - Thijs Simonian on <t:1754672400:F>. Be sure to submit any questions you have here.
minor tide
@everyone <@&1372208590248742964> We are sharing a new dataset with over 140k conversations from the text arena collected between April 17th and July 25th 2025. Join us as we explore real-world trends, new features, and fresh prompts.
What’s covered in the latest analysis:
- Language & topic breakdowns
- Rating changes: How Arena scores shift over time
- Overview of the released dataset
And more!
The data analysis highlights not just who is winning, but why, and what signals might matter most in human-based AI evaluations.
Read the full breakdown here on our blog.
minor tide
<@&1372208635530448926> <@&1398740297521037332> New model capability added to Video Arena
- Veo 3 Fast & Veo 3 now has **Image-to-Video with audio **capabilities!
Give it a try using /image-to-video in our video-arena channels: #video-arena-1 #video-arena-2 #video-arena-3 and vote on what you think is best!
minor tide
<@&1372208635530448926> New models added to Text & WebDev Arena on LMArena!
- OpenAI gpt-oss-120b
- OpenAI gpt-oss-20b
- Claude Opus 4.1 (battle mode only)
minor tide
<@&1372208635530448926> <@&1372208524230397962> <@&1398740297521037332> Thanks to the contributions of this community we now have Video Leaderboards available!
minor tide
<@&1372208635530448926> <@&1398740297521037332> New models have been added to Video Arena
- Hailuo-02-pro
- Hailuo-02-fast
- Sora
- Runway-Gen4-turbo
Give them a try in our video-arena channels: #video-arena-1 #video-arena-2 #video-arena-3
minor tide
@everyone Now that OpenAI's GPT-5 is here, we're thrilled to share that this model has been setting a new bar across our leaderboards. Tested under the codename “summit”, GPT-5 now holds the highest Arena score to date.
Thanks to the community's voting GPT-5 is now number #1 in Text, Vision, and WebDev Arena.
GPT-5 is now available on LMArena!
minor tide
<@&1398740297521037332> Reminder that tomorrow we're having our Staff AMA focussing on Video Arena. If you have questions please fill out this form.
https://discord.com/events/1340554757349179412/1400149736027328623
minor tide
<@&1372208635530448926> New models added to LMArena!
- gpt-5-mini-2025-08-07
- gpt-5-nano-2025-08-07
minor tide
<@&1398740297521037332> Reminder we'll be starting this event in 10 minutes!
https://discord.com/events/1340554757349179412/1400149736027328623
minor tide
Thank you all for being a part of this amazing community. Whether you've been here since day one or recently joined us, the LMArena team is incredibly grateful that you're all here!
minor tide
<@&1372208635530448926> New models added to Search Arena on LMArena!
- gpt-5-search
- claude-opus-4.1-search
minor tide
<@&1385704581677191278> Thank you for your patience with this one! Vote on July's contest submissions here! On Friday 8/15 we'll announce the winner and start our next contest.
minor tide
<@&1372208635530448926> <@&1372208524230397962> Thank you all for patiently waiting for this update. We're happy to share the leaderboards were just updated including the GPT-5 variants.
- gpt-5-high
- gpt-5-chat
- gpt-5-mini-high
- gpt-5-nano-high
minor tide
<@&1385704581677191278> To celebrate the launch of our experimental Video Arena let's generate some videos into our August contest!
How does it work and what are the rules?
- Use
/videoto generate two videos related to our contest theme (Slice 🔪 ) in our video arena channels (#video-arena-1) .To submit your generation to the contest -> **Forward the generated message to the #august-contest channel ** This is where we are collecting submissions.
- To forward a message, hover your mouse over the message, select Forward, send to the #august-contest channel.
- Submissions will close on Sept 12th!
What could I win?
- 1 Month of Discord Nitro
- Become the newest member to receive the <@&1378032433873555578> role!
August's Contest Theme - Slice! 🔪 Show us those oddly satisfying, safe‑for‑work, crisp cross‑section cuts into everyday objects. Think is it cake videos. Examples here & here.
Big congrats to @delicate iris for being our July <@&1378032433873555578> ! Check out their generation here.
minor tide
@everyone <@&1372208635530448926> We're excited to announce that we're partnering with DataTecnica LLC & the National Institutes of Health to create BiomedArena! This new arena is focussed on real-world biomedical workflows, from literature review to disease modeling, using open, reproducible methods trusted by scientists.
It's already in use at NIH’s Intramural Research Program. We're proud to support and help expand this work around:
- Open, reproducible evaluations
- Expert-in-the-loop feedback
- Scientific transparency at scale
Check out BiomedArena for yourself here and read more about this partnership on our blog here!
minor tide
<@&1398740297521037332> We’re introducing #highlight ! You’ll now notice a ⭐ has been added to the area you vote in Video Arena. This ⭐ allows users to **highlight spectacular generations they want shown off! ** After 4 people chose to highlight a generation it’ll automatically be posted to #highlight for all to appreciate.
Reminder we have our August Contest currently running, be sure to forward your 🔪Slice 🔪 generations in #august-contest - more info here.
minor tide
<@&1372208635530448926> - Our legacy website began this wonderful journey of exploring the world's leading AI models and shaping community leaderboards.
Looking toward the future, we've decided to invest our full efforts into the current version of the site. We have bittersweet news to share: our legacy site is no longer available. All the great features from the legacy site are either under consideration or currently in development. We're listening—please tell us in our feedback forum which features matter most to you.
A heartfelt thanks from the LMArena team to everyone who's been on this journey with us from the early stages! The legacy site will always have a place in our hearts. ❤️
minor tide
<@&1372208635530448926> - New model added to Image Edit on LMArena!
- qwen-image-edit
minor tide
<@&1372208524230397962> - The Image Edit Leaderboard has been updated and Qwen-Image-Edit is now the #1 open model for Image Edit!
Check out our leaderboards here.
minor tide
<@&1372208635530448926> New models added to LMArena!
- deepseek-v3.1
- deepseek-v3.1-thinking
minor tide
<@&1398740297521037332> Thank you for your patience while we resolved the issue. The @royal stirrup Bot is working again.
Reminder how to use the bot -> type /video or /image-to-video only in these channels #video-arena-1 #video-arena-2 #video-arena-3
minor tide
<@&1372208473269473320> We'd love to understand better why you use your preferred version.
Please fill out this survey if you'd like to share your thoughts. 
minor tide
@everyone <@&1372208635530448926> A lot of you have found a recent anonymous model a-peeling & today we're thrilled to share with you that nano-banana = Gemini-2.5-Flash-Image-Preview 🍌
The previous two weeks of testing this model has led to the largest Elo score jump in LMArena history. The text-to-image leaderboard & image-edit leaderboard have been updated and a new leader of the pact has emerged.
Gemini-2.5-Flash-Image-Preview is now available in Battle, Side by Side, & Direct modes. Try it out now & join us in #nano-banana to tell us what you think!
minor tide
<@&1372208635530448926> - A new model provider has landed on our text leaderboard. Microsoft AI's MAI-1-preview is now sitting at #13!
Come check out MAI-1-preview available now on LMArena.
minor tide
You all have been waiting patiently for this feature and we’re thrilled to share that **we're starting to roll out User Login! ** On our canary site you can currently login with your Google Account 
.
A few notes:
- You will be able to access your Chat History on different devices when logged in.
- This is currently only available on our canary site. Desktop & mobile.
- When you create/login to an account, you can merge your existing chats with the account with the
Merge existing chats with your accounttoggle.- To log out open the sidebar, locate your email on the bottom-left, click the three dots.
We want to ensure this feature is working properly which is why we’re slowly rolling this out. Make us aware of any bugs in here: #1343291835845578853 and be sure to share any feedback in here: #1372230675914031105 . Plans for more sign-in options are being worked on.
minor tide
@everyone <@&1372208635530448926> You all have been waiting patiently for this feature and we’re thrilled to share that **we're starting to roll out User Login! ** You can currently login with your Google Account .
A few notes:
- You will be able to access your Chat History on different devices when logged in.
- When you create/login to an account, you can merge your existing chats with the account with the
Merge existing chats with your accounttoggle.- To log out open the sidebar, locate your email on the bottom-left, click the three dots.
- There is a small hold-out group, meaning some users won't yet have access to this feature yet. We will roll this out to everyone as soon as we're certain everything is working properly.
Be sure to share any bugs in here: #1343291835845578853 and be sure to share any feedback in here: #1372230675914031105 . Plans for more sign-in options are being worked on.
minor tide
<@&1398740297521037332> Reminder that you only have 9 days left to submit your for our Video Gen Contest!
How does it work and what are the rules?
- Use
/videoto generate two videos related to our contest theme (Slice 🔪 ) in our video arena channels (#video-arena-1) .- To forward a message, hover your mouse over the message, select Forward, send to the #august-contest channel.
July's Contest Theme - Slice! 🔪 Show us those oddly satisfying, safe‑for‑work, crisp cross‑section cuts into everyday objects. Think is it cake videos. Examples here & here.
minor tide
<@&1398740297521037332> The bot is working again! Thank you all for your patience while we worked on a fix.
Reminder how to use the bot:
- You must be in #video-arena-1 , #video-arena-2 , or #video-arena-3
- Type
/video- Type in your prompt and hit
Enter
minor tide
<@&1372208635530448926> – Due to unprecedented traffic, we’re introducing rate limits for image generation. Logged-in users will continue to enjoy higher limits, and we’ll keep making the login experience better so contributing to community evaluations becomes even more rewarding.
You can learn more about User Login here.
minor tide
<@&1372208635530448926> New Models added to LMArena!
- Qwen3-max-preview
- Kimi-K2-0905-preview
minor tide
@everyone Multi-turn editing is now available on all image edit models! Refine your image step by step instead of trying to fit every edit into one mega-prompt.
Multi-Turn for Image Edit is available in Battle, Side by Side, or Direct. Try it out for yourself.
minor tide
<@&1398740297521037332> Due to increased usage of the experimental Video Arena, we're going to make a change setting the individual use limit to 5 generations per day. After you've hit this limit, you will have to wait 24hr to start using the bot again.
Reminder how to use Video Arena can be found here.
minor tide
<@&1372208635530448926> New Model added to LMArena!
- Seedream-4
minor tide
<@&1372208635530448926> New Models added to LMArena!
- Qwen3-next-80b-a3b-instruct
- Qwen3-next-80b-a3b-thinking
minor tide
<@&1372208635530448926> New Model added to LMArena!
- Hunyuan-image-2.1
minor tide
<@&1372208635530448926> New Model added to LMArena!
- Seedream-4-high-res
minor tide
<@&1372208419418673302> <@&1372208243681660978> <@&1372207111445938226> We'd love to understand better why you use your preferred version.
Please fill out this survey if you'd like to share your thoughts.
minor tide
<@&1385704581677191278> Thank you to everyone that participated in our first Video Arena GenAI contest! Vote on which you like the best to crown our new <@&1378032433873555578>
This contest theme is 🔪Slice🔪 Show us those oddly satisfying, crisp cross‑section cuts into everyday objects.
minor tide
<@&1372208524230397962> Our Text-to-Image & Image Edit leaderboards have been updated with some interesting movement.
Seedream-4-high-res is now tied with Gemini-2.5-flash-image-preview (nano-banana) for the #1 slot on the Text-to-Image leaderboard. Check it out yourself on the Text-to-Image leaderboard Here.
For Image Edit we're now seeing Seedream-4-high-res holding the #2 position. Image Edit leaderboard can be found Here.
Tell us what you think in our #leaderboards channel.
minor tide
<@&1372208635530448926> It's our mission to improve the reliability of AI. We're introducing an evaluation product to analyze human-AI interactions at scale, turning their complexity into insights the AI ecosystem will benefit from.
Our AI Evaluation service offers enterprises, model labs, and developers comprehensive evaluations grounded in real-world human feedback. LMArena AI Evaluations consist of:
- Comprehensive, in-depth evaluations based on feedback from our community.
- Auditability through representative samples of feedback data.
- Service-level agreements (SLA) with committed delivery timelines for evaluation results.
Analytics based on community feedback reveal strengths, weaknesses, and tradeoffs—helping providers build even better models and AI applications for everyone.
Are you an enterprise, model lab, or developer that wants to learn more about our AI Evaluation services? Read more on our blog.
minor tide
@everyone <@&1372208524230397962> New open models have entered the Text Arena, and the rankings by provider have shifted for September. Only the top 7 open models also rank within the top 50 overall (proprietary & open).
Qwen-3-235b-a22b-instruct is currently in the top slotLongcat-flash-chat debuts on the charts in impressive manor landing at #5 Holding Firm
Qwen-3-235b-a22b-instruct stays at #1 (overall rank #8)Kimi-K2-0711-preview firm at #2 (overall rank tied for #8)DeepSeek-R1-0528 holding steady at #3 (overall rank #9)GLM-4.5 holds at #4 (overall rank #13)Mistral-Small-2506 at #9 (overall rank tied at #53) New Entrants
Longcat-flash-chat debuts at #5 (overall rank #20) Movers
MiniMax-M1 went from #5 → #6 (overall rank #43)Gemma-3-27b-it shifts from #6-> 7 (overall rank #46)gpt-oss-120b drops to #8 (overall rank #51)Llama-3.1-Nemotron-Ultra-253b-v1 drops from #8 -> #10 (overall rank #53) Dropouts
Command-A-03-2025 (#10 → out)Check out the details for yourself on our leaderboards. Let us know what you think in the #leaderboards channel.
minor tide
@everyone <@&1372208635530448926> New models added to LMArena!
- Grok-4-fast
- Grok-4-fast-search
Grok-4-fast-search by xAI was tested under the codename menlo and has rocketed as #1 on the Search Leaderboard! Text Arena also tested Grok-4-fast, codename tahoe, where it has debuted impressively at #8 on the Text Leaderboard.
Check out the rankings for yourself and let us know what you think in #leaderboards.
minor tide
<@&1372208635530448926> Thank you all for your patience with us as we made adjustments to seedream-4. An update has been made where seedream-4-2k is now available in Battle, Direct, & Side by Side. The model known as seedream-4-high-res is not available at this time. Keep an eye on this channel as we'll continue to provide new model updates in this channel.
Give seedream-4-2k a try here!
minor tide
<@&1372208635530448926> New Models added to LMArena!
- deepseek-v3.1-terminus
- deepseek-v3.1-terminus-thinking
minor tide
<@&1372208635530448926> New Models added to LMArena!
- qwen3-max-2025-09-23
- qwen3-vl-235b-a22b-thinking
- qwen3-vl-235b-a22b-instruct
minor tide
<@&1372208635530448926> New Models added to LMArena's WebDev!
- Gpt-5-codex
- Qwen3-coder
Give them a try and vote here!
minor tide
<@&1372208635530448926> <@&1372208524230397962> Seedream-4-2k has landed on the leaderboards! On the Text-to-Image leaderboard at #1, Seedream-4-2k is now tied with Gemini-2.5-flash-image-preview (nano-banana)! On the Image Edit leaderboard Seedream-4-2k is now ranked at #2.
Let us know what you think in #leaderboards!
minor tide
<@&1372208635530448926> New Models added to LMArena!
- gemini-2.5-flash-preview-09-2025
- gemini-2.5-flash-lite-preview-09-2025
minor tide
<@&1372208635530448926> - New model added to WebDev on LMArena!
- claude-sonnet-4-5-20250929
Give it a try here!
minor tide
<@&1372208635530448926> - New models added to LMArena!
- claude-sonnet-4-5
- claude-sonnet-4-5-20250929-thinking-16k
minor tide
<@&1372208635530448926> - New models added to LMArena!
- deepseek-v3.2-exp
- deepseek-v3.2-exp-thinking
minor tide
We are deeply grateful that you all are interested in being a part of this community. From the LMArena Team - thank you all!
minor tide
<@&1372208635530448926> New model added to LMArena!
- glm-4.6
minor tide
<@&1385704581677191278> We want to see what you all can make with LMArena! Could you be crowned our next <@&1378032433873555578> ? October's AI Gen Contest is now open.
How does it work and what are the rules?
- You must submit your entry by sharing a screenshot in #october-contest
- On October 24th submissions will be closed and we'll share a way to vote.
- Submissions must be done through Battle Mode
- Your submission must include both the left and right response.
- Your submission must be after you've voted for which response you prefer meaning the models should be revealed in your submission.
- Example here.
- 1 month of Discord Nitro
- <@&1378032433873555578> role! You’ll hold onto this exclusive role which will be hoisted towards the top of the member’s list.
Use wild shapes, vibrant colors, and chaotic lines to express feelings or ideas. Create a unique visual experience. This month’s contest will be for image creations only.
Congrats to @vernal lance for being our first Video Gen <@&1378032433873555578>
minor tide
@everyone We've thrilled to share with you all the Arena Champions Role! Our goal with this role, <@&1422628364782407830> , is to create a better community for those interested in in-depth AI discussions. This program aims to reward members who show genuine commitment to meaningful conversation by providing a private space where they can engage without interruptions.
Access to this space will be granted through an application process. Members must demonstrate both interest in AI and commitment to meaningful conversation. If you're looking for a dedicated space to have these discussions...
As a thank you to longtime community members, we've granted automatic access to those who've been part of this server since July 2025. Note that members with this role will need to Follow the Category to view these new channels. You can find this in the Channels & Roles tab at the top of the channel list. Select Browse Channels, then enable the Arena Champions Category. You'll then be able to see the list of channels!
minor tide
<@&1372208635530448926> Reasoning Trace is now available on Side by Side & Direct chat with reasoning models. Think of this as a way for reasoning models to show their work before they provide a response. Check it out in Side by Side & Direct now!
<@&1372208635530448926> New model added to LMArena!
- reve-v1
Note this model is** image-edit only**, meaning it will only work if you upload an image for it to edit. The model will error out when using text-to-image.
New model update as well. This has replaced the 16k version.
- claude-sonnet-4-5-20250929-thinking-32k
minor tide
<@&1372208635530448926> <@&1372208524230397962> Our Text Leaderboard has been updated!
Claude Sonnet 4.5 has made it onto the Text Leaderboard, impressively tied with Claude Opus 4.1 for the #1 slot. It’s also shining across many other categories, including: Hard Prompts, Coding, Creative Writing, Instruction Following, and others.
Share your thoughts with us in #leaderboards
minor tide
<@&1372208635530448926> New model added to LMArena!
- ibm-granite-h-small (ibm)
minor tide
<@&1398740297521037332> New model added to LMArena's Video Arena!
- ray-3
Reminder on how to use Video Arena can be found here: #1397655624103493813
minor tide
@everyone As we continue to build LMArena it’s important we remain focussed we’re delivering the tools you all need to excel at being knowledge experts. In order to do this we must understand what is important to you all better.
If you’re interested in sharing your expertise with the team please…
minor tide
<@&1398740297521037332> <@&1372208635530448926> New models added to LMArena's Video Arena!
- sora-2
- sora-2-pro
note these are only going to be available in text-to-video
Reminder on how to use Video Arena can be found in #1397655624103493813
minor tide
<@&1372208635530448926> New models added to LMArena!
- hunyuan-vision-1.5-thinking
- ring-flash-2.0
- ling-flash-2.0
minor tide
Introducing #codename-discussion We'll use this channel to have focussed discussions related to models that are using codenames. Reminder that these models appear under codenames or aliases in Battle mode.
You may have to enable this channel manually in
Channels & Roles->Browse Channels
minor tide
Fill Out This Survey
Apply Here
minor tide
<@&1372208524230397962> <@&1372208635530448926> Sora 2 & Sora 2 Pro have now landed on the Text-to-Video Leaderboard! Sora 2 Pro is now tied as #1 alongside Veo 3 & Veo 3 Fast. Sora 2 has also shaken things up by landing in at #3. If you haven’t already, be sure to check them out for yourself in our Video Arena! Reminder for how to can be found here: #1397655624103493813
Be sure to let us know what think in #leaderboards !
minor tide
<@&1372208635530448926> New models added to LMArena!
- qwen3-vl-8b-thinking
- qwen3-vl-8b-instruct
minor tide
<@&1372208635530448926> New model added to LMArena & WebDev!
- claude-haiku-4-5-20251001
https://x.com/arena/status/1978523872505872481
New models added to Video Arena!
- veo-3-1-fast
- veo-3-1
X
minor tide
<@&1372208635530448926> <@&1372208524230397962> The Text Leaderboard has been updated! Claude-Haiku-4-5 has landed and is currently sitting in the #22 rank. Be sure to check out the Text Arena Leaderboard and let us know what you think in #leaderboards
minor tide
<@&1372208635530448926> <@&1372208524230397962> There has been a big shift in our Text-to-Video Leaderboard & Image-to-Video Leaderboard as Veo-3.1 now ranks #1 in both!
Haven’t tried it out for yourself yet? Be sure to check out #1397655624103493813 for an explanation on how Video Arena works.
Let us know what you think in #leaderboards and be sure to share those Veo-3.1 generations in #ai-creations
X
minor tide
<@&1372208635530448926> - New model added to LMArena!
- minimax-m2-preview
https://x.com/arena/status/1981850766039187901
X
minor tide
<@&1372208635530448926> <@&1372208524230397962> We’ve updated the Image-to-Video Leaderboard. Hailuo-2.3 is now on the leaderboard and is ranked #5 with Seedance-v1-pro & Kling-2.5-turbo-1080p.
minor tide
<@&1372208635530448926> <@&1372208524230397962> New model added to LMArena's Video Arena! This is an image-to-video model only.
- hailuo-2.3-fast
Also, the Text-to-Video Leaderboard has been updated! Hailuo-2.3 is now ranked #7. Check it out and let us know what you think in #leaderboards.
minor tide
<@&1385704581677191278> Thank you to everyone that participated! Vote on which you like the best to crown our new <@&1378032433873555578>.
This contest theme is 🎨 Abstract Art 🎨 Use wild shapes, vibrant colors, and chaotic lines to express feelings or ideas. Create a unique visual experience.
minor tide
<@&1372208635530448926> <@&1372208524230397962> A new model has landed on the WebDev Leaderboard—MiniMax-M2 is now the #1 top open model, & top #4 overall. The community has shown that it shines at performance coding, reasoning, and agentic-style tasks while remaining cost-effective and fast.
minor tide
@everyone We’re thrilled to introduce a new tagging system built on our evaluation framework that’ll identify the most expert-level prompts from the community. With this new system we’re introducing Expert Leaderboard. Arena Expert reveals the structure of prompts: their depth, reasoning, and specificity, which drives the clarity of evaluation.
Additionally, Occupational Leaderboards are now available that map prompts to real-world domains. By mapping all Arena prompts across 23 occupational fields, the system captures the full spectrum of real-world reasoning tasks. With this update you’ll see 8 of these leaderboards live including:
minor tide
<@&1372208635530448926> - New model added to LMArena!
- kimi-k2-thinking
https://x.com/arena/status/1986482438768673107
X
minor tide
<@&1372208635530448926> Ernie-5.0-preview-1022 is now on the Text Arena Leaderboard and is now in the #2 rank! Let us know what you think in #leaderboards
minor tide
<@&1372208635530448926> The Image Edit Leaderboard has been updated! Reve-edit-fast is now publicly released and is now ranked in the top 5.
Check out the Image Edit Leaderboard yourself!
minor tide
<@&1385704581677191278> Congrats to @dusty stump for being our October's Abstract Art Contest Winner!! The newest member of our <@&1378032433873555578> ! Check out their generation here.
Stay tuned for future contest announcements.
minor tide
<@&1372208635530448926> <@&1372208524230397962> The Text leaderboard has been updated and Kimi-k2-thinking is now the the #2 ranked open source ranked model & tied for #7 overall. We’ve been seeing this model excel at Math, Coding, and Creative Writing categories. On our Expert leaderboard, Kimi-k2-thinking has an impressive score of 1447 as well.
Check out the Text leaderboard yourself and let us know what you think in #leaderboards.
minor tide
@everyone Driven by community feedback, User Login with email is now available! Save your chat history across multiple devices on both your mobile and desktop browsers.
minor tide
@everyone Code Arena is now available on LMArena! The WebDev Arena has leveled up with a complete redesign shaped by community feedback and is now known as Code Arena. With Code Arena, models generate live, deployable web apps and sites that anyone can open, inspect, and judge directly, in real time.
Since Code Arena’s evaluation methods have been rebuilt, a fresh new leaderboard designed to reflect this new system has launched.
Try out Code Arena for yourself - HERE
Check out the new Code Arena leaderboard - HERE
Learn more in our Blog Post - HERE
minor tide
<@&1372208635530448926> - New model added to Text, Vision, and Code Arena on LMArena!
X
minor tide
<@&1372208524230397962> <@&1372208635530448926> Today we're announcing an important update to how model rankings are displayed on LMArena, one that makes them both more interpretable and more statistically accurate in how they reflect uncertainty. There will now be two new metrics displayed alongside each model’s score:
- Raw Rank: the model’s position based purely on its Arena score.
There are no ties here. Each model receives a unique rank based on its performance.- Rank Spread: an interval that shows the range of possible ranks a model could have, given the overlap in confidence intervals (CIs) across models.
To learn more about this update check out this blog post here>) and let us know what you think in #leaderboards.
minor tide
<@&1372208635530448926> - New model added to Text & Vision!
- gpt-5.1-high
& new models added to Code Arena!- gpt-5.1-codex
- gpt-5.1-codex-mini
minor tide
<@&1372208524230397962> <@&1372208635530448926> The Text Arena leaderboard has been updated! Grok-4.1-thinking is now in the #1 rank and followed by Grok-4.1 in the #2 rank. On the Expert leaderboard, Grok-4.1-thinking is also in the #1 rank excelling at Hard Prompts, Coding, Instruction Following, and Creative Writing.
Check out the Text Arena leaderboard and bookmark our Leaderboard Changelog for leaderboard updates.
minor tide
<@&1385704581677191278> We want to see what you all can make with LMArena! To celebrate Code Arena's launch let's use this modality! Could you be crowned our next <@&1378032433873555578> ? November's AI Gen Contest is now open.
How does it work and what are the rules?
- You must submit your entry by sharing the preview link in #november-contest
- On
November 28thDecember 10 submissions will be closed- Example here
- 1 month of Discord Nitro
- <@&1378032433873555578> role!
Looking for more information on how to use Code Arena? Be sure to check out our walkthrough video here !
minor tide
<@&1372208524230397962> <@&1372208635530448926> The Image-to-Video & the Text-to-Image Leaderboards have been updated as well. Wan2.5-i2v-preview & Wan2.5-t2i-preview have landed in the Top 5 on the Image-to-Video and Text-to-Image leaderboards.
Check out the Image-to-Video leaderboard & the Text-to-Image leaderboard! Let us know what you think in #leaderboards.
Stay up to date with our Leaderboard Changelog!
minor tide
<@&1372208524230397962> <@&1372208635530448926> Gemini-3-pro has landed on the Text, WebDev, and Vision leaderboards!
- #1 in Text scoring 1501
- #1 in Vision scoring 1328
- #1 in WebDev scoring 1487
You can try out Gemini-3-pro for yourself as it’s now available on LMArena! Be sure to bookmark our Leaderboard Changelog for all leaderboard related updates.
minor tide
<@&1372208524230397962> <@&1372208635530448926> A new model provider has entered the WebDev Arena: Deep Cogito has released Cogito-v2.1 which ties ranks #18 overall and is in the Top 10 for Open Source models!
See for yourself on the WebDev Leaderboard and share your thoughts in #leaderboards
minor tide
<@&1372208524230397962> <@&1372208635530448926> Today, some scores for GPT-5.1 are live for the Text Arena:
GPT-5.1-high ranks #4GPT-5.1 ranks #12Stay tuned as we collect more votes and see how the scores converge for GPT-5.1-high in the new WebDev leaderboard powered by Code Arena. We’ll also see how GPT-5.1-medium stacks up.
minor tide
<@&1372208635530448926> - Google DeepMind’s new image model just landed on LMArena.
- gemini-3-pro-image-preview (nano-banana-pro)
X
minor tide
<@&1372208524230397962> <@&1372208635530448926> A new model provider has landed on the Vision leaderboard! Ernie-5.0-preview-1022 by Baidu debuts with a score of 1206!
Check out the Vision leaderboard and share some prompts you’ve used with Ernie-5.0-preview-1022 in #share-prompts
minor tide
<@&1372208524230397962> <@&1372208635530448926> The WebDev leaderboard has been updated.
GPT 5.1’s Code Arena evaluations have been added:
GPT-5.1-mediumlanded at #2 with a score of 1407GPT-5.1landed at #8 with a score of 1364GPT-5.-Codexlanded at #9 with a score of 1336GPT-5.1-Codex-Minilanded at #13 with a score of 1252
Check out the WebDev leaderboard and contribute with your votes on Code Arena. Always stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208524230397962> <@&1372208635530448926> It’s been a major week for leaderboard movement, and now the Text-to-Image and Image Edit leaderboards have added Gemini-3-pro-image-preview.
Gemini-3-pro-image-previewranks #1 on the Text-to-Image leaderboard (+84 pt over nano-banana)Gemini-3-pro-image-previewranks #1 on the Image Edit leaderboard (+41 pt over nano-banana)
Let us know what you think in #leaderboards.
minor tide
<@&1372208635530448926> - New model added to Text and Code Arena on LMArena!
- claude-opus-4-5-20251101
https://x.com/arena/status/1993035224880759068
X
minor tide
<@&1372208635530448926> Driven by community feedback, we've updated how multi-turn for image edit works and added some new features we’re excited to share with you:
Edit feature in image generation.Big shoutout to everyone in the community who spoke up when we first launched multi-turn for image generation. Your feedback means a lot!
minor tide
<@&1372208635530448926> - New model added to Text-to-Image and Image Edit on LMArena!
- flux-2-pro
- flux-2-flex
https://x.com/arena/status/1993444903876280645
X
minor tide
<@&1372208635530448926> - New models added to Search Arena!
- gemini-3-pro-grounding
- gpt-5.1-search
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Claude-opus-4-5-20251101 & Claude-opus-4-5-20251101-thinking-32k have been added to the leaderboards!
WebDev leaderboard (powered by Code Arena)
- #1 for Claude-Opus-4.5 (thinking-32k)
- #2 for Claude-Opus-4.5
Expert leaderboard- #1 for Claude-Opus-4.5
Text leaderboard- #3 for Claude-Opus-4.5
- #6 for Claude-Opus-4.5 (thinking-32k)
Stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208635530448926> - New models added to Text Arena!
- deepseek-v3.2
- deepseek-v3.2-thinking
https://x.com/arena/status/1995564824718442620
X
minor tide
<@&1372208524230397962> <@&1372208635530448926> - new models Flux-2-pro, Flux-2-flex, and KAT-coder-pro-v1 have landed on the leaderboards!
- Flux-2-flex ranks #3
- Flux-2-pro ranks #5
Image Edit leaderboard- Flux-2-pro ranks #6
- Flux-2-flex ranks #7
WebDev leaderboard- KAT-coder-pro-v1 ranks #16
Always stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - A new open source model has landed on the leaderboard.
Mistral-Large-3 lands at #6 among open models and #28 overall on the Text leaderboard. Mistral-Large-3 was tested under the codename “Jaguar” and performs strongly in:
- Coding
- Hard Prompts
- Multi-Turn
- Instruction Following
- Longer Query
Check out the Text leaderboard for yourself, and let us know what you think in #leaderboards
minor tide
<@&1372208635530448926> Resharing an Early Access Program that we haven't mentioned in a bit!
The LMArena Test Garden is our private feedback program that'll invite selected members to get a sneak peeks at features, design mocks, and ideas the team is considering implementing for early feedback. For those who are exceptional at providing feedback, and would like to see what we're cooking up, this program is for you!
Note that if selected we'll followup privately for next steps. Being apart of this program does require signing of an NDA.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Search Arena leaderboard has been updated!
- Gemini-3-pro-grounding ranks #1
- Gpt-5.1-search ranks #2
Tell us what you think in #leaderboards and stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208635530448926> - New model added to Text Arena!
- nova-2-lite
https://x.com/arena/status/1996396395411177920
X
minor tide
<@&1372208524230397962> <@&1372208635530448926> - New model added to Text-to-Image Arena & Image Edit Arena!
- Seedream-4.5
The leaderboards for Text-to-Image & Image Edit have been updated to include Seedream-4.5 as well! This model has landed at #3 for Image Edit leaderboard and ranks #7 in Text-to-Image leaderboard!
Stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208635530448926> - New models added to Code Arena & Video Arena!
- Gpt-5.1-codex-max
Video Arena- Kling-2.6
X
minor tide
<@&1385704581677191278> - Reminder, our current Code Arena contest is going to wrap up on December 10th! Be sure to add those submissions to #november-contest before time runs out.
More details here.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated and ERNIE-5.0-Preview-1103 landed with a score of 1431 putting it in the top 20.
Check out the Text Arena leaderboard and stay up to date with the Leaderboard Changelog.
minor tide
<@&1385704581677191278> - Our November Code Arena Contest is now closed!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - New models added to Code Arena & Text Arena!
- GPT-5.2-high
- GPT-5.2
Tested internally under the codename “robin and robin-high”, GPT-5.2-high now ranks #2 & GPT-5.2 now ranks #6 on our WebDev leaderboard! These scores are preliminary, so stay tuned as they stabilize. Always stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208635530448926> - New models added to Text and Vision Arena!
- glm-4.6v
- glm-4.6v-flash
minor tide
@everyone - We recently launched our YouTube channel! If you’ve been enjoying LMArena, you’ll want to subscribe: we’re posting fast, practical breakdowns to help you understand the AI frontier and choose the best models for your work.
Recent videos include:
• Beginner’s guide to free + open models
• GPT-5.2 enters the Arena
• Why small open models are disappearing (7B → 32B shift)
• Generating SVGs to measure coding capabilities
• How to choose the best AI model for coding
Subscribing helps us grow it, and ensures you see new releases the moment they drop. Let us know if there are topics you want us to cover!
minor tide
<@&1385704581677191278> We want to see what you all can make with LMArena! Could you be crowned our next <@&1378032433873555578> ? December's AI Gen Contest is now open.
How does it work and what are the rules?
- You must submit your entry by sharing a screenshot in #december-contest
- On December 30th submissions will be closed and we'll share a way to vote
- Submissions must be done through Battle Mode
- Your submission must include both the left and right response
- Your submission must be after you've voted for which response you prefer meaning the models should be revealed in your submission
- Example here
What could I win?
- 1 month of Discord Nitro
- <@&1378032433873555578> role! You’ll hold onto this exclusive role which will be hoisted towards the top of the member’s list
We'd love to see what you created on LMArena! Everyone is encouraged to post your contest submission on your own X account and tag @arena. We may repost what you've shared!
Let's get festive! We want to see how you celebrate the holidays with diverse celebrations like Christmas, Hanukkah, Kwanzaa, New Year's and more. This month’s contest will be for image creations only.
Help me give a big congrats to @quiet rivet for being our first Code Arena <@&1378032433873555578> !! Check out the winning submission here.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text to Image leaderboard and Image Edit leaderboard have new models shaking up the ranks!
gpt-image-1.5 is #1 in Text-to-Image (1264)chatgpt-image-latest is #1 on Image Edit (1409)gpt-image-1.5 #4 in Image Edit (1395)New models added to Image Arena!
- gpt-image-1.5
- chatgpt-image-latest
minor tide
<@&1372208524230397962> <@&1372208635530448926> - GPT-5.2-high has landed on the Text Arena leaderboard at #13 in the Text Arena!
With a score of 1441, the model performs strongest in:
Stay up to date with all leaderboards changes with our Leaderboard Changelog!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Gemini-3-flash has landed on our Leaderboards! The Text Arena leaderboard, Vision Arena leaderboard, and WebDev Arena leaderboard have all been updated.
Gemini-3-Flash highlights:
Where Gemini-3-Flash (thinking-minimal) performs strongest:
These models are now available on Text and WebDev Arena.
- gemini-3-flash
- gemini-3-flash (thinking-minimal)
Let us know what you think in #leaderboards and stay up to date with our Leaderboard Changelog.
sly rock
@everyone - Today we’re releasing Arena-Rank, an open-source Python package for paired-comparison ranking—the same code that powers the LMArena leaderboards.
Why we’re doing this:
Under the hood, Arena-Rank reflects several methodological and engineering upgrades we’ve made over the past months, including faster JAX-based optimization and cleaner separation between data preprocessing and modeling.
We’re looking forward to feedback, experiments, and contributions from the community. We can't wait to see what kind of leaderboards you make.
💬 To install → pip install arena-rank
This is part of our broader commitment to open science and transparent AI evaluation—and it’s just a starting point. Read more in our blog here.
sly rock
<@&1372208524230397962> <@&1372208635530448926> - New models reve-v1.1and reve-v1.1-fast have landed on the leaderboard!
This represents a +6-point gain over Reve V1.
Always stay up to date with our Leaderboard Changelog.
<@&1372208524230397962> <@&1372208635530448926> - The [Search Arena leaderboard](<https://lmarena.ai/leaderboard/search) has been updated. GPT-5.2-Search ranks #2 while Grok-4.1-Fast-Search ranks #4.
Both models debuted ahead of their predecessors, posting gains of +10 points for GPT-5.2-Search and +17 points for Grok-4.1-Fast-Search.
Stay up to date with our Leaderboard Changelog and let us know what you think about how GPT-5.2 ranks in #leaderboards.
sly rock
<@&1372208524230397962> <@&1372208635530448926> - The [Text Arena leaderboard](<https://lmarena.ai/leaderboard/text) has been updated. GPT-5.2 makes its debut and ranks #17.
Compared to GPT-5.1, the model has improved by +2 points. It trails just one point behind GPT-5.2-high, which is optimized for expert-level reasoning and critical tasks.
Stay up to date with our Leaderboard Changelog!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated & ERNIE-5.0-Preview-1203 by Baidu has landed with a score of 1451. Here are some highlights:
ERNIE-5.0-Preview-1103Bookmark our Leaderboard Changelong to stay up to date with the latest changes to our leaderboards!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Our WebDev leaderboard has been updated and GLM-4.7 by Z.ai ranks #6. This makes it the new #1 open model for WebDev. GLM-4.7 has a score of 1449, which is a +83 pt increase over GLM-4.6.
Try out GLM-4.7 in Code Arena and share with the community some of the generations made in #share-prompts !
minor tide
<@&1372208635530448926> - New model added to Video Arena!
- seedance-v1.5-pro
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Our Text Arena leaderboard has been updated with GLM-4.7 & Minimax-m2.1-preview included.
Let us know what you think in #leaderboards , and stay up to date with all leaderboards changes with our Leaderboard Changelog.
minor tide
<@&1385704581677191278> - Our December Contest is now closed!
minor tide
<@&1372208635530448926> - New models added to Image Arena & Image-Edit Arena!
- qwen-image-2512
- qwen-image-edit-2511
https://x.com/arena/status/2007273636512837958
X
minor tide
<@&1372208635530448926> - Over the break, we identified some issues with the user login and registration flow. These have now been fixed, so if you were experiencing problems, please try logging in or registering again. If you continue to have any issues, don’t hesitate to let us know in #1451836386293448725
Many thanks to everyone who reported these problems initially, we really appreciate your help!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image Edit leaderboard has been updated and qwen-image-edit-2511 is now the #1 open model, and #9 overall!
On the Text-to-Image leaderboard qwen-image-2512 is the #2 open model, and ranks #13 overall.
Find our Leaderboard Changelog here and share what you think about these updates in #leaderboards.
minor tide
<@&1385704581677191278> We want to see what you all can make with LMArena! Could you be crowned our next <@&1378032433873555578> ? For January we're going to be running a contest each week!
How does it work and what are the rules?
- You must submit your entry by sharing a screenshot in #january-1st-contest
- On January 9th submissions will be closed and we'll share a way to vote
- Submissions must be done through Battle Mode
- Your submission must include both the left and right response
- Your submission must be after you've voted for which response you prefer meaning the models should be revealed in your submission
- Example here
What could I win?
- 1 month of Discord Nitro
- <@&1378032433873555578> role! You’ll hold onto this exclusive role which will be hoisted towards the top of the member’s list
Create an image of something that represents you looking out the window and looking torwards your wildest, brightest future. Make it aesthetic, surreal, or sci-fi! This month’s contest will be for image creations only.
Shoutout to @meager wren for being our new <@&1378032433873555578> !! Check out the winning submission here.
sly rock
@everyone
Today, we're excited to announce our $150M funding round at a post-money valuation of more than $1.7B, nearly triple our valuation just seven months after our seed raise in May. The round was led by Felicis and UC Investments, with participation from a16z, The House Fund, LVDP, Kleiner Perkins, Lightspeed and Laude Ventures. This milestone reflects a growing industry consensus: AI cannot scale responsibly without independent, transparent, and continuous evaluation.
LMArena started as a research experiment. It’s now becoming a foundational pillar for the AI ecosystem. To this community who has tested, voted, reported bugs, submitted suggestions, and shared your perspective: Thank you. You are shaping the future of AI. ♥️
Let’s measure and advance what the world needs next. We will move even faster to build new features and improve our product experience for this community to evaluate the frontier of AI.
Read more about the announcement on our blog: https://news.lmarena.ai/series-a/
▶️ A message just for our wonderful community:
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Vision Arena leaderboard has been updated. ERNIE-5.0-Preview-1220 is now ranked #8 with a score of 1226. Baidu currently stands as the only Chinese lab in the Top 10 on the Vision leaderboard.
Check out our Leaderboard Changelog for all leaderboard updates.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Hunyuan-Video-1.5is now on our leaderboards! For the Text-to-Video leaderboard Hunyuan-Video-1.5 now ranks #18 with a score of 1193 and on the Image-to-Video leaderboard it ranks #20 with a score of 1202.
Let us know what you think in #leaderboards and always stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard just got an update. Let us know what you think in the #leaderboards channel.
minor tide
<@&1372208635530448926> - A new model has been added to Video Arena.
- ltx-2-19b
Check out #1397655624103493813 and test outltx-2-19bnow!
minor tide
<@&1385704581677191278> We want to see what you all can make with LMArena! Could you be crowned our next <@&1378032433873555578> ? For January we're going to be running a contest each week!
How does it work and what are the rules?
- You must submit your entry by sharing a screenshot in #jan
- On January 16th submissions will be closed and we'll share a way to vote
- Submissions must be done through Battle Mode
- Your submission must include both the left and right response
- Your submission must be after you've voted for which response you prefer meaning the models should be revealed in your submission
- Example here
What could I win?
- 1 month of Discord Nitro
- <@&1378032433873555578> role! You’ll hold onto this exclusive role which will be hoisted towards the top of the member’s list
This month’s theme :
🍃 Nature Reclaims… depictions of a world where nature has begun to reclaim what humanity once built or occupied. Show us human-made environments overtaken, transformed, or reinterpreted by the natural world. 🍃
Vote for our January's First Contest Winner! Vote Here.
minor tide
<@&1372208635530448926> - New models have been added to Video Arena.
- veo-3.1-audio-4k
- veo-3.1-audio-1080p
- veo-3.1-fast-audio-4k
- veo-3.1-fast-audio-1080p
Check out #1397655624103493813 to try them out yourself!
minor tide
<@&1372208635530448926> - A new model has been added to Code Arena!
- gpt-5.2-codex
&
New model has been added to Image Arena!- glm-image
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated! ERNIE-5.0-0110 now ranks #8 with a score of 1460 along with being #12 in Arena Expert. This is currently the only model from a Chinese lab in the Top 10. It performs strongest in the Math category and quite a few occupational categories.
Try out Text Arena and stay up to date with changes in our leaderboards with the Leaderboard Changelog.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Image Arena leaderboard has been updated where z-image-turbo now ranks #22, flux.2-klein-9B now ranks #24, and flux.2-klein-4B ranks #31 overall. Additionally, Image Edit Arena leaderboard has been updated where flux.2-klein-9B ranks #15 and flux.2-klein-4B ranks #21.
Stay up to date with our Leaderboard Changelog!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image Edit leaderboard has been updated. wan2.5-i2i-preview has been added and now ranks #21 with a score of 1213.
As always, stay up to date with our Leaderboard Changelog.
minor tide
<@&1385704581677191278> We want to see what you all can make with LMArena! Could you be crowned our next <@&1378032433873555578> ?
How does it work and what are the rules?
- You must submit your entry by sharing the **Code Arena preview link ** in #january-3rd-contest
- On January 26th submissions will be closed and we'll share a way to vote
- Submissions must be done with Code Arena. Reminder on how to use Code Arena can be found here.
- Example here.
What could I win?
- 1 month of Discord Nitro
- <@&1378032433873555578> role! You’ll hold onto this exclusive role which will be hoisted towards the top of the member’s list
This month’s theme :
⌨️ Code Arena - Let's use Code Arena for this contest! No specific theme! Build what ever you think would be most appealing!
Big congrats to @dark glacier for being our first January contest winner! Check out their submission here.
Vote for our January's Second Contest Winner! Vote Here. Reminder of the theme - Nature Reclaims… 🍃
minor tide
Text Arena has officially passed 5 million community votes. That’s millions of real-world comparisons shaping how frontier AI models are evaluated. You didn’t just prompt. You tested. You voted. You moved the leaderboard.
This milestone belongs to you. 💙
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Image Arena leaderboard has been updated and GLM-Image now ranks #8 among open models and #35 overall with a score of 1018.
minor tide
@everyone What started last summer as a small Discord bot experiment has grown into a rigorous way to measure and understand how frontier video models perform with real-world use. Thank you to our wonderful community for all the feedback! Today, Video Arena is now available to all on LMArena.
Learn more about Video Arena on our blog here.
minor tide
<@&1398740297521037332> <@&1372208635530448926> - Have you tried Video Arena on web yet? Generate videos with 15 different frontier AI models and compare them head-to-head. Vote for the best output to power the leaderboards.
Get the full walkthrough and a few pro-tips from one of our lead engineers on our YouTube channel: https://www.youtube.com/watch?v=jaIU6eKVK1M
minor tide
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image Edit Arena leaderboard now has more data around real-world use. It now has two distinct leaderboards:
- Leader change: ChatGPT Image (Latest) goes #1 -> #3, while Gemini 3 Pro Image 2K (Nano‑Banana Pro) goes #2 -> #1.
- Biggest rise: FLUX-2-Flex jumps #19 -> #12 (up 7 places).
- Small‑model mover: FLUX-2-Klein 4B climbs #22 -> #17 (up 5 places).
- Biggest drops: Seedream-4 2K slides #7 -> #14 (down 7 places) and Qwen Image Edit (2511) slips to #11 -> #16 (down 5 places)
Check it out yourself to see the differences on our Image Edit leaderboard.
minor tide
<@&1372208635530448926> - New models added!
Text-to-Image
- wan2.6-t2i
Image Edit- wan2.6-image
Code Arena- devstral-2
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena.
- qwen3-max-thinking
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image Edit leaderboard has been updated. Hunyuan-Image-3.0-Instruct now ranks #7 for Image Edit.
Try out Hunyuan-Image-3.0-Instruct vs. all the best frontier models in Image Arena. Stay up to date with our Leaderboard Changelog.
minor tide
minor tide
minor tide
<@&1372208635530448926> - A few reminders for everyone:
minor tide
<@&1372208635530448926> - Auto-Modality & Model selector are now live!
Auto-Modality: Whether your prompt is a coding question, a math proof, or an image generation request, auto-modality routes it to the right modality automatically. This makes every evaluation smoother and more intuitive. For more information on how auto-modality works check out this Help Center article.
Model selector: A new design for our model selection window is now live in Direct and Side by Side. Models are now ordered by rank and can be filtered by modality. Find more information about this in the Model Selection Menu Help Center article
minor tide
<@&1398740297521037332> - Check out our Better AI videos in under 90 seconds video now on our Youtube channel: https://www.youtube.com/watch?v=0hCI2XEh0x0
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated. Kimi K2.5 Thinking is now ranked the #1 open model and ranking #15 overall.
Some highlights:
Kimi K2.5 Thinking has also been added to Code Arena so go check it out.
minor tide
@everyone - We’re excited to share today our new look and feel to match our scientific mission: to measure and advance the frontier of AI for real-world use. We are now just: Arena. Now available at: arena.ai.
From a small PhD research project to a platform powered by a global community of millions. This rebrand has been shaped by this community, the people who use it.
Read more about the rebrand process on our blog here.
minor tide
@everyone - As our community continues to grow, it’s important to keep conversations organized and easy to follow.
Introducing the #ask-here channel. This channel is the** home for one-off questions**. Going forward, questions in #general will be discouraged so discussions there can stay focused on AI related discussions.
For reporting issues please use #1343291835845578853. Before posting, check whether a thread already exists and add your report there.
For sharing feedback please use #1372230675914031105. As with bugs, add to an existing thread if one is already active.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Vision Arena leaderboard has been updated. Kimi-k2.5-thinking is now the #1 open model and ranks #6 overall in Vision Arena making it the only open model in the Top 15.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The leaderboards have been updated with new models being added! Check them out:
Text-to-Image leaderboard
- p-image
- wan2.6-t2i
- p-image-edit
- wan2.6-image
Text-to-Video leaderboard
- kling-o1-pro
- ltx-2-19b
- ltx-2-19b
Code Arena leaderboard
- devstral-2
Text Arena leaderboard
- glm-4.7-flash
Search Arena leaderboard
- gemini-3-flash-grounding
- claude-sonnet-4-5-search
- claude-opus-4-5-search
- Gpt-5.2-search-non-reasoning
Stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208635530448926> - Two new features have rolled out to everyone.
Search Bar - Your chats are now searchable, with the option to filter by modality.
Archive Chat - Archive chat sessions to keep them for later without cluttering your chat history.
With these features now live, the process for deleting a chat session has changed. Follow the steps in this article to learn how to delete chats going forward.
minor tide
<@&1398740297521037332> - Video Arena on Discord had it's rate limit updated to 1 generation request per 24 hour period. Using Video Arena on web still has the same rate limit of 3 generations per 24 hour period.
minor tide
<@&1372208635530448926> - New models have been to Arena!
Text Arena
- step-3.5-flash
Code Arena- qwen3-max-thinking
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard has received an update. Kimi-K2.5-thinking now ranks #1 open and #5 overall on Code Arena.
The community has also ranked Kimi-K2.5-thinking as #1 open model for Vision, and Text including the Coding category.
Let us know what you think in #leaderboards and share the previews of what you’ve built with Kimi.ai in #ai-creations .
minor tide
@everyone - Max is Arena’s intelligent router, powered by 5+ million real-world community votes.
Max routes each prompt to the most capable model with latency in mind. AI models excel at different things (code, math, speed, reasoning). Max orchestrates across model strengths to deliver reliable performance across real-world use cases.
To learn more about Max check out this blog post and our Youtube video.
minor tide
minor tide
<@&1372208635530448926> - Max intelligently routes each prompt to the most capable model currently live on Arena.
Catch the full video on our YouTube with Arena researcher Derry.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image-to-Video leaderboard has been updated. Vidu-Q3-pro by Vidu AI is now in the Top 5 with a score of 1362.
minor tide
<@&1372208635530448926> - New models added to Text Arena and Code Arena.
- claude-opus-4-6
- claude-opus-4-6-thinking
minor tide
<@&1372208635530448926> - Our AI Capabilities Lead, Peter, breaks down the latest performance of Opus 4.6. Check it out here.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Claude Opus 4.6 has landed on our leaderboards and is now #1 across Code, Text and Expert!
#1 in Code Arena: +106 score over Opus 4.5
#1 in Text Arena: scoring 1496, 10pts over Gemini 3 Pro and is also ranking #1 in key Text Arena categories:
#1 for Expert Arena: +~50 leadTell us what you think about these changes in #leaderboards and stay up to date with our Leaderboard Changelog.
minor tide
<@&1385704581677191278> - Thank you all for voting for our 2nd January AI Generation Contest 🍃 Nature Reclaims! The votes have been tallied and the newest member of our <@&1378032433873555578> is @raw thunder ! Check out the winning submission here.
Help crown our next <@&1378032433873555578> by voting. Reminder that this theme is: ⌨️ Code Arena
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Kimi K2.5 is now on our leaderboards and is in the top 5 open models for Vision, Text, and Code!
#2 open model in Vision, #10 overall on par with gpt-5.1
#3 open model in Text, #26 overall on par with o3 and Qwen3-max-preview
#4 open model in Code, #10 overall rivaling gemini-3-flash
minor tide
<@&1398740297521037332> - We’re making a small but important change: Video Arena is moving off Discord and will be exclusively available on arena.ai.
This change was driven largely by community feedback requesting new features that aren’t possible to support through a Discord bot. Moving Video Arena to our site gives us the flexibility to build and ship new capabilities that Discord simply can’t support. While the Discord version is going away, Video Arena itself is not . It’s still fully available on our site and will continue to improve over time.
Starting Wednesday, February 11th at 4pm PST, Video Arena will no longer be available through Discord. The Video Arena site experience on arena.ai/video is unaffected and will remain fully accessible.
Thank you everyone for being vocal about how you want to see Video Arena improve. This move helps us build a better experience for everyone.
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Image and Image Edit leaderboards have been updated to include Grok-Imagine-Image.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard and the Code Arena leaderboard have been updated to include `Claude-opus-4-6-thinking’!
Some highlights:
minor tide
@everyone - Text-to-image models have advanced quickly, and so have use cases. After analyzing 4M+ user prompts (from fantasy art to logos and posters), it’s clear that a single leaderboard is no longer enough to capture real-world use. With that in mind, we’ve updated the Text-to-Image Arena with Prompt Categories & Quality Filtering.
Category-specific leaderboards that surface domain-level performance across common use cases. New categories include:
• Product, Branding & Commercial Design
• 3D Imaging & Modeling
• Cartoon, Anime & Fantasy
• Photorealistic & Cinematic Imagery
• Art
• Portraits
• Text Rendering
To improve reliability, we filtered the prompt set to focus on inputs that consistently deliver quality image generation. After removing ~15% of noisy or underspecified prompts, we recomputed the leaderboard, resulting in more stable, higher-confidence rankings.
These updates are a first step toward more granular, interpretable evaluation of text-to-image models—grounded in how people actually use them. You can now explore how your favorite text-to-image models perform across these categories on the Text-to-Image Arena leaderboard.
minor tide
<@&1398740297521037332> - Reminder that on **Wednesday February 11th @ 4pm PST ** the Video Arena through the Discord bot will not be available. Video Arena will still be available through the site and is unaffected by this change. This shift allows us to focus efforts into improving Video Arena with features and capabilities that aren't possible through a Discord bot.
We appreciate everyone who has provided feedback and enjoyed using Video Arena through Discord. Thank you!
minor tide
<@&1372208590248742964> <@&1372208635530448926> - Today we’re announcing our Academic Partnerships Program, a new initiative to support independent academic research in AI evaluation, rankings, and measurement.
As AI systems advance and adoption accelerates, the methods we use to evaluate and compare models increasingly shape both scientific progress and real-world outcomes. Many of the most important contributions in this area come from the academic research community, and we’re proud to help support that work directly.
Selected projects may receive up to $50,000 in research funding. We welcome proposals across evaluation methodology, leaderboard design, measurement and statistical validity, preference data and human evaluation, and safety/alignment evaluation.
Learn more about our Academic Partnerships Program here.
Apply to our Academic Partnerships Program by March 31, 2026 here.
minor tide
<@&1372208635530448926> - Upload PDFs with your prompts to add richer context and test models on document reasoning, bringing evaluations closer to real-world use. Try it across 10 models today - we’ll be adding more over time.
Leaderboard coming soon. Start uploading, comparing, and voting!
minor tide
<@&1372208635530448926> <@&1398740297521037332> <@&1372208524230397962> - The Video Arena leaderboards have been updated and high-res 1080p variants for Veo 3.1 now rank #1 and #2 in Video Arena.
In Text-to-Video the 1080p versions top the chart
- #1 veo-3.1-audio-1080p
- #2 veo-3.1-fast-audio-1080p
In Image-to-Video, 1080p variants make the top 5
- #2 veo-3.1-audio-1080p
- #5 veo-3.1-fast-audio-1080p
New models have been added to Video Arena and Text Arena.
- veo-3.1-audio-1080p (Video Arena)
- veo-3.1-fast-audio-1080p (Video Arena)
- step-3.5-flash (Text Arena)
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena and Code Arena.
- glm-5
minor tide
<@&1372208635530448926> - Since launching Code Arena in November to evaluate frontier AI models on real-world, agentic coding tasks, we’ve received a lot of feedback asking to adapt more complex workflows.
With multi-file apps, you can now build and compare production-ready projects, making it easier to evaluate how top frontier AI models perform on your actual use cases.
minor tide
<@&1372208635530448926> <@&1372208524230397962> - The Text Arena leaderboard has been updated and glm-5 is now #1 among open models.
Stay up to date with our Leaderboard Changelog.
minor tide
<@&1398740297521037332> - We are currently in the process of removing the Video Arena through the Discord bot. Video Arena will still be available through the site and is unaffected by this change. This shift allows us to focus efforts into improving Video Arena with features and capabilities that aren't possible through a Discord bot.
We appreciate everyone who has provided feedback and enjoyed using Video Arena through Discord. Thank you!
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena and Code Arena.
- Minimax-m2.5
minor tide
<@&1372208635530448926> <@&1372208524230397962> - The Code Arena leaderboard has been updated and GLM-5 is now the #1 open model in Code Arena. Overall #6 on par with Gemini-3-pro, 100+pts below Claude-Opus-4.6 in agentic webdev tasks.
Arena's AI Capability Lead Peter Gostev shares his first impressions of two powerful models: GLM-5 and MiniMax-M2.5. Give it a watch here.
minor tide
<@&1372208635530448926> - A new model has been added to Text, Vision, and Code Arena.
- qwen3.5-397b-a17b
minor tide
<@&1372208635530448926> - A new model has been added to Text, and Code Arena.
- claude-sonnet-4-6
minor tide
<@&1473460945308221542> - Check our newest YouTube video with Arena's AI Capability Lead Peter Gostev sharing his first impressions of Claude Sonnet 4.6, Anthropic's latest model in the Claude family.
https://www.youtube.com/watch?v=b0yr1I0dxA4
Want the new YouTube Updates role? Just head to Channels & Roles (in the channel list), click Customize, choose What brings you here, and select YouTube Update.
minor tide
<@&1372208635530448926> - New models have been added to Search Arena.
- sonnet-4.6-search
- opus-4.6-search
minor tide
@everyone - Millions of votes power the leaderboard. Now you can filter for what matters to you. A new side panel lets you filter and break down ranked results to find the best model for your task. Some highlights:
Filter by category (e.g. Coding, Expert prompts)
Open vs. Proprietary Models
Rank labs by their top-performing models
Check it out, and let us know what you think in #leaderboards.
minor tide
<@&1473460945308221542> - Join our Designer, Justin Keoninh, for a walkthrough of the new leaderboard UI updates and learn how to make the most of the latest enhancements.
https://www.youtube.com/watch?v=xfmcR6-Uh5Q
Want the new YouTube Updates role? Just head to **Channels & Roles **(in the channel list), click Customize, choose What brings you here, and select YouTube Update.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated including Qwen3.5-397B-A17B.
In the highly competitive Text Arena, a few highlights:
#20 overall in Text on par with Claude Opus 4.1 variants
Top 5 open for key categories in Text like: Math, Instruction Following, Multi-Turn, Creative Writing and Coding
Top 5 for open models in Arena Expert (#26 overall)
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena and Code Arena leaderboards have been updated and now include Gemini-3.1-Pro. It’s top 3 across Text and Vision Arena, and #6 in Code Arena, tied closely with Claude Opus 4.5.
Highlights:
Tied #1 in Text (scoring 1500) only 4 pts from Opus 4.6
Top 3 in Arena Expert (scoring 1538), just behind Opus 4.6
#6 in Code Arena, on par with Opus 4.5 and GLM-5
minor tide
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard and Text Arena leaderboard has been updated to include Claude-sonnet-4.6.
Highlights:
minor tide
<@&1398740297521037332> - We’re planning to remove the Video Arena generation channels from the server on Monday 2/23 @ 4pm PST. If you’d like to download any generations, please make sure to do so before that date.
<@&1473460945308221542> - Ever wondered what actually happens after you vote on Arena? Clayton breaks down the full journey.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Vision Arena leaderboard has been updated to include Qwen3.5-397B-A17B. It's now tied for top 2 open model in the Vision Arena with Kimi-K2.5-Instant. Ranks #13 overall on par with proprietary models like GPT-4o.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated to include GPT-5.2-chat-latest now in the top 5.
Highlights include:
Top 5 scoring 1478 on par with Gemini-3-Pro
+40pt improvement over the GPT-5.2 model
Top in key categories: Multi-Turn, Instruction-Following, Hard Prompts, Coding
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image Arena leaderboard has been updated to include Reve V1.5.
Highlights:
#4, scoring 1177, on par with Grok-Imagine-Image
Top 5 for categories: Text Rendering, Art and Product, Branding Commercial Design
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard has been updated now including Qwen3.5-397B-A17B.
Highlights:
top 7 open model
ranks #17 overall, on par with proprietary models like GPT-5.2 and Gemini-3-Flash
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Video leaderboard and the Image-to-Video leaderboard has been updated and now include Wan2.6-t2v.
Highlights:
Wan2.6-t2v is the #1 Chinese model in the Video Arena
Top 8 for Text-to-Video, scoring 1346, on par with Veo-3-fast-audio
#12 for Text-to-Image, scoring 1292, close with Seedance v1.5 pro and Kling 2.6 pro
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Search Arena leaderboard and Text Arena leaderboard has been updated and now include Grok-4.20-Beta1.
Highlights:
#1 in Search Arena, scoring 1226, leading GPT-5.2 and Gemini-3
#4 in Text Arena, scoring 1492 on par with Gemini 3.1 Pro
minor tide
<@&1372208635530448926> - New models have been added to Code, Text, and Vision Arena.
Code Arena
- qwen3.5-27b-code
- qwen3.5-35b-a3b-code
- qwen3.5-122b-a10b-code
Text and Vision Arena- qwen3.5-27b
- qwen3.5-35b-a3b
- qwen3.5-122b-a10b
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image Edit Arena leaderboard has been updated. Seedream-5.0-Lite now ties for top 5 on the Multi-Image Edit Arena.
Highlights:
ranks #10 in Single-Image, scoring 1301 on par with Hunyuan-Image-3.0 and Nano Banana
ranks #23 overall for Text-to-Image Arena, scoring 1106
minor tide
<@&1372208524230397962> <@&1372208635530448926> - P-Video enters the Video Arena leaderboards in top 26.
Highlights:
tied for #22 in Text-to-Video, score 1178 on par with Hailuo 2.0 and Kandinsky 5.0 Pro
top 26 for Image-to-Video, score 1199 on par with Hailuo 2.0 Fast
their fastest model with pricing at $0.04/second for 1080p
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Nano Banana 2 debuts at #1 in Image Arena, and it changes the game again 🍌Officially released as Gemini-3.1-Flash-Image-Preview, it introduces a new web search capability, unlocking image generation grounded in real-world context.
Highlights:
#1 Text-to-Image scoring 1279, surpassing GPT-Image-1.5 and Nano Banana Pro
Ties for #1 Single-Image Edit, scoring 1407 on par with ChatGPT-Image-Latest
Top 3 Multi-Image Edit, alongside Nano Banana Pro variants
$0.067 per image ~2x cheaper than Nano Banana Pro
sand grail
<@&1473460945308221542> - Peter covers three reasons AI agents underperform inside existing software.
https://www.youtube.com/watch?v=qF8afKUGRpc
Want the new YouTube Updates role? Head to Channels & Roles (in the channel list), click Customize, choose What brings you here, and select YouTube Updates.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Search Arena leaderboard has been updated to include Claude-Opus-4-6 and Claude-Sonnet-4-6.
Highlights:
#1 wide lead by Opus 4.6 scoring 1255, +30pt over Grok-4.20-beta1, GPT-5.2 and Gemini-3
Sonnet 4.6 ranks #7 on par with GPT-5.1
minor tide
<@&1372208635530448926> - A new model has been added to Code Arena!
- gpt-5.3-codex
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Video Arena leaderboard has been updated to include Kling-V3-Pro.
Highlights:
tied #8, scoring 1337 on par with Wan2.5-i2v-preview
+52pt improvement over Kling 2.6 Pro
+48pt over Kling-2.5-turbo-1080p
sand grail
<@&1473460945308221542> Guanglei Song, PhD introduces 7 new categories in Image Arena to find the top models for photorealistic, 3D modeling, and more.
sand grail
<@&1473460945308221542> Arena in 60 seconds. What did we miss?
sly rock
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Video Arena leaderboard has been updated to include Runway Gen 4.5.
Gen-4.5 scores 1218, on par with KlingAI’s Kling-2.6-Pro.
sly rock
<@&1372208524230397962> <@&1372208635530448926> - The leaderboards have been updated to include Gemini-3.1-Flash-Lite-Preview for Text and Code Arena.
Highlights:
ranks #36 in Text, scoring 1432, on par with Grok-4.1-fast, strong in Creative Writing, and Longer Query
surpassing larger Gemini 2.5 Flash and GPT-5-mini
tied in Code Arena for #35 scoring 1261, on par with Qwen3-coder for agentic webdev tasks
sand grail
<@&1473460945308221542> - in Document Arena, you can upload a PDF and watch two anonymous AI models go head-to-head.
sly rock
<@&1372208524230397962> <@&1372208635530448926> - The Document Arena leaderboard has been added. The Document Arena displays model rankings based on side-by-side evaluations of real-world document reasoning performance across user-uploaded PDF files.
See which frontier AI models rank highest in document reasoning, all powered by side-by-side evaluations on user-uploaded PDFs from real work use cases.
#1 is Claude Opus 4.6 scoring 1525, +51 pts in the lead
While Opus 4.5 and Gemini 3.1 Pro Preview join in the top 3
Latest GPT-5.2 tied at #9, ~100 pts behind Opus 4.6
minor tide
<@&1372208635530448926> - New models have been added to Text Arena and Video Arena!
- GPT-5.3-Chat-Latest (Text Arena)
- PixVerse V5.6 (Video Arena)
sand grail
<@&1473460945308221542> - Arena ML researchers Derry and Evan go behind the scenes of Arena's new Max intelligent router.
https://www.youtube.com/watch?v=nO6E5t6dmA0
Want the new YouTube Updates role? Head to Channels & Roles (in the channel list), click Customize, choose What brings you here, and select YouTube Updates.
sly rock
<@&1372208524230397962> <@&1372208635530448926> - The Text & Code leaderboards have been updated to include Qwen 3.5 medium models: qwen3.5-27b ,qwen3.5-35b-a3b , qwen3.5-122b-a10b and qwen3.5-flash
Code Arena Highlights:
top 10 open Qwen3.5-122b-a10b, scoring 1384 and Qwen3.5-27b, scoring 1375 both very close to proprietary models: Claude Sonnet 4.5 and GPT-5.1-medium
Qwen3.5-35b-a3b, scoring 1257 is on par with new Gemini-3.1-flash-lite-preview
Qwen3.5-Flash, scoring 1243 is on par with GPT-5.1-codex-mini
Text Arena Highlights:
top 10 open Qwen3.5-122b-a10b, scoring 1420
Qwen3.5-27b, the smallest and densest scores 1410, on par with GLM-4.5
Qwen3.5-35b-a3b, scoring 1392 and Qwen3.5-Flash scoring, 1395 are on par with 6-7x larger last-generation model Qwen3-235b
sly rock
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated to include gpt-5.4 and gpt-5.4-high.
Highlights:
GPT-5.4-high is tied with Gemini-3-Pro.
Top 3 in Creative Writing, and top 10 in Instruction Following, Hard Prompts.
Top 6 for Occupational categories - Writing, Literature & Language, Entertainment, Sports & Media, Business, Management & Financial Ops.
GPT-5.4 (reasoning none) ranks #16
sand grail
<@&1473460945308221542> - AI capability lead Peter Gostev runs through one-shot tests to see how GPT 5.4 compares to other models.
https://www.youtube.com/watch?v=foEfcttIuiI
Want the new YouTube Updates role? Head to Channels & Roles (in the channel list), click Customize, choose What brings you here, and select YouTube Updates.
sand grail
<@&1473460945308221542> - no words, just visuals.
sly rock
<@&1372208524230397962> <@&1372208635530448926> - The Video Arena leaderboards have been updated to include pixverse-v5.6
Highlights:
#15 on Text-to-Video
#15 on Image-to-Video
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Document Arena leaderboard has been updated to include claude-sonnet-4-6
#2 ranking overall
top 3 are all Anthropic models
sand grail
<@&1473460945308221542> - After testing GPT 5.4-medium and coming away underwhelmed, Peter revisits the model at higher reasoning levels — and the difference is massive.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated and now includes Nemotron-3-Super-120B-A12B.
Highlights:
#37 open rank in Expert
#38 open rank in Math
#118 Text overall
minor tide
sand grail
<@&1473460945308221542> - Every LLM can retrieve sources but the real challenge is reasoning about which ones to trust → arena.ai/search
minor tide
<@&1372208524230397962> <@&1372208635530448926> - GPT-5.4 lands tied #2 on Document Arena and in top 5 for Arena Expert. In the Document Arena, top models for document analysis and long-form reasoning are ranked based on real-world use.
Highlights:
#2 tied with Sonnet 4.6
#5 for Arena Expert
top 10 in Business, Management, & Financial Ops and Writing, Literature, & Language
top 15 in Math, Instruction Following, Multi-Turn & Hard Prompts
top 15 in Text Arena overall
sand grail
<@&1473460945308221542> - Are you old enough to remember the Nano Banana hype? Meet the engineer who added it to the Arena - Yue!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard has been updated and GPT-5.4-high has landed in the top 6.
Highlights:
top 6 in Web Dev overall
#6 for Multi-File React
top 10 for Single-File HTML
sand grail
<@&1473460945308221542> - Peter takes us through his viral benchmark 💩
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard and the Code Arena leaderboard has been updated to include Grok 4.20 Beta Reasoning.
Highlights:
#7 in Text Arena overall tied with GPT-5.4-high
top 10 in Math, Multi-Turn, Creative Writing, Coding & Hard Prompts
top 15 in Expert Arena
minor tide
<@&1372208524230397962> - Everyone's real-world use for AI differs. Select the columns and data that matters most to you:
minor tide
<@&1372208524230397962> - Today we’re launching the Video Edit Arena to evaluate the frontier capability of video models! The leaderboard is powered by thousands of real-world community votes. Click the Edit button in Video Arena to edit any video and compare top model outputs. More models coming soon!
#1 Grok-Imagine-Video
#2 Kling-o3-pro
#3 Kling-o1-pro
#4 Gen4-aleph
minor tide
<@&1372208635530448926> - New models have been added to Text Arena & Vision Arena!
- gpt-5.4-mini-high
- gpt-5.4-nano-high
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena and Code Arena!
- minimax-m2.7
sand grail
<@&1473460945308221542> - Peter explains why the most underrated quality in AI models isn't how smart they are - it's how consistently they perform.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard has been updated to include MiniMax-M2.7
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard and the Expert Arena leaderboard has been updated to include Qwen3.5-max-preview.
Highlights:
#3 Math
#10 Expert
#15 Text Arena
Top 20 for Writing, Literature & Language, Life, Physical, & Social Science, Entertainment, Sports, & Media, and Medicine & Healthcare
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image Arena leaderboard has been updated to include MAI-Image-2.
Highlights:
#5 in Text-to-Image overall
#5 for 3D Imaging & Modeling, Cartoon, Anime & Fantasy, Photorealistic & Cinematic Imagery, Art and Portraits
#6 for Product, Branding & Commercial Design
The Vision Arena leaderboard has also been updated to include Grok-4.20-Beta-Reasoning
Highlights:
Scoring 1240
#11 across all Vision
minor tide
<@&1372208635530448926> - We’ve recently updated our bug reporting process. If you encounter a Something went wrong error, you’ll now see a Trace ID. We recommend the following steps when this happens:
minor tide
<@&1372208635530448926> - A new model has been added to Vision Arena!
- mimo-v2-omni
sand grail
<@&1473460945308221542> - three ways to test AI - each designed for a different use case.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard and Arena Expert leaderboard has been updated to include MiMo-V2-Pro.
Highlights:
top 6 lab, #13 in Code Arena for agentic webdev tasks
#10 for Arena Expert
top 20 for Life, Physical, & Social Science and Business, Management, & Financial Ops occupational categories
sand grail
<@&1473460945308221542> - curious about the math behind Arena's ranking system? This is a technical deep dive into the core methodology with Anastasios Angelopoulos, co-founder and CEO of Arena.
minor tide
<@&1372208635530448926> - As our community continues to grow, we’re making a few updates to help us provide better and more organized support. Going forward, here are some updated guidelines:
The #general channel should be used for general AI discussion. If you have a question, please use #ask-here. If you’re reporting a bug, please use #1343291835845578853. Going forward, I wouldn't expect getting an answer from staff if it's asked in general.
If you’d like to reach the team directly, please send a message to @hearty vessel. Direct messages to individual team members (including myself) will no longer be supported.
Before creating a new thread in #1343291835845578853, please check if there’s already an existing thread for the issues you're experiencing. Duplicate posts may be removed to help keep things organized.
sand grail
<@&1473460945308221542> - how to create an Arena account to unlock higher rate limits, cross-device chat sync, and access additional features.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Search Arena leaderboard has been updated to include Gemini-3.1-Pro-Grounding.
sand grail
<@&1473460945308221542> - We looked at 3 years of Arena data. Proprietary models still hold the top 20 spots — but open source is climbing — with the top open source models ranked 20 (GLM-5 by Z.ai), 23 (Kimi-K2.5-Thinking by Moonshot AI) and 27 (Qwen3.5-397b-a17b by Alibaba).
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated to include GPT-5.4-Mini-High.
Highlights:
#3 Business, Management & Financial Ops
#10 Multi-Turn
#13 Arena Expert
#17 Legal & Government
#19 Instruction Following
sand grail
<@&1473460945308221542> Arena ML researchers Evan and Derry discuss whether the future of AI belongs to massive generalist models or smaller, fine-tuned specialists.
sand grail
<@&1473460945308221542> Five things to try if you see a "Something went wrong with this response" error and how to submit a bug report if you need extra help.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard, Vision Arena leaderboard, and Search Arena leaderboard has been updated to include Grok-4.20-multi-agent-beta-0309.
Highlights:
#7 for Search Arena, #11 in Text Arena, #22 in Vision Arena
#3 Medicine & Healthcare
#6 Expert Prompts
#6 Mathematical
#6 Legal & Government
minor tide
<@&1372208524230397962> <@&1372208635530448926> - We've added Pareto frontier charts to the leaderboard! The Pareto frontier curve demonstrates which models are most efficient at their level of performance (by Arena score) vs. a blended price per 1M tokens (3:1 Ratio).
Now available across:
Text Arena
Vision Arena
Search Arena
Document Arena
Code Arena
In this video, Peter and Justin walk through how to read the Pareto frontier, find hidden gems, compare models across categories.
minor tide
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard has been updated to include Qwen 3.6 Plus. Qwen 3.6 Plus Preview is the #2 lab for the React leaderboard in Code Arena which ranks models based on agentic workflows involving multi-step reasoning, tool use, and multi-file apps.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated to include Gemma-4-31B.
Highlights:
#3 open (#27 overall), on par with the best open models Kimi-K2.5, Qwen-3.5-397b
Top 3 across Math, Instruction Following, Multi-Turn, Hard Prompts, Creative Writing, and Coding
Apache 2.0 license
Its efficient variant: Gemma-4-26B-A4B is #6 open (#39 overall)
minor tide
<@&1372208524230397962> <@&1372208635530448926> - We're releasing the full history of Arena leaderboard data as a public dataset, nearly 3 years of rankings across 10 Arenas, dozens of categories, and hundreds of models. Optimized to empower analysis and unlock new insights across modalities and over time.
Learn more in our blog post here.
Find the dataset on our Hugging Face here.
minor tide
<@&1372208635530448926> - We are changing some of the models available in Direct and Side-by-Side modes. These changes are part of our efforts to ensure that we can continue offering access to AI models while keeping the platform running reliably.
The following models are being removed from Direct and Side-by-Side:
- claude opus models
- gpt 5.4, gpt-5.4-high
- gemini-3.1-pro-preview
Updates like this help us maintain availability for everyone over the long term, and we intend to bring these models back in a way that’s more sustainable when possible.
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena & Code Arena!
- Qwen-3.6-Plus
sand grail
<@&1473460945308221542> - Longer context windows, more decisive votes, and over 90% correlation with traditional battle rankings — plus new signals about human preference that weren't measurable before.
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena and Code Arena!
- GLM-5.1
sand grail
<@&1473460945308221542> - GLM-5.1 is a solid incremental upgrade over GLM-5, with slightly higher richness and quality across most generations. But the gap isn't massive.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated to include GLM-5.1.
Highlights:
#1 open model in Longer Query (#4 overall)
#1 open model in Life, Physical & Social Science (#5 overall)
#1 open model in Entertainment, Sports & Media (#8 overall)
#1 open model in Coding (#10 overall)
sand grail
<@&1473460945308221542> - This is one of the longest-running datasets tracking real model performance over time.
minor tide
<@&1372208635530448926> - We're sharing an early opportunity to share feedback at a change we’re planning before we release anything. But before introducing the change, we'd like to provide some context. Right now, Arena manages usage through a mix of per-model limits, session token limits, and modality limits. It works, but it's difficult to track, and we’re confident we can build something better.
We're reworking this into a single, unified system. Instead of separate hidden limits everywhere, you'd have a daily balance of “credits” you can spend however you want, so you can go all-in on Video Arena, stick with one model, bounce between a few, whatever. The goal is to give you way more control and visibility over how you use Arena .
We’re still early, so many details are intentionally flexible. Things like credit structure, launch timing, and even the final name are still being refined. We’re sharing this now so feedback like yours can directly shape how it evolves. If you’re wondering what certain aspects will look like, there’s a good chance we’re actively working through those decisions and won't have a direct answer.
Drop your thoughts in this thread - https://discord.com/channels/1340554757349179412/1491461236448170134 What would make this feel fair? What would you want to know up front? What would actually be useful to see?
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard has been updated to include GLM-5.1.
sand grail
<@&1473460945308221542> Derry and Peter break down why verifiable benchmarks only capture a narrow slice of model performance, and why human judgment still matters more than most people think.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard and the Vision Arena leaderboard has been updated to include Muse-Spark.
Highlights:
Text Arena: #3
Vision Arena: #2
#4 Hard Prompts, #6 Coding, #9 Creative Writing, #10 Instruction Following, #27 Expert
#3 tied for Business, Management, & Financial Ops, #7 Legal & Government, #12 Writing & Literature
sand grail
<@&1473460945308221542> - ICYMI check out Peter's keynote from the AI Engineer conference in London last week.
sand grail
<@&1473460945308221542> Can AI function like a professional Accountant? We asked an expert to judge AI responses to a complex Arena Expert-style prompt.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Video Edit Arena leaderboard has been updated to include Happyhorse-1.0. HappyHorse-1.0 by Alibaba-ATH debuts at #1 in Video Edit Arena!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Image to WebDev Leaderboard ranks models based on their ability to generate websites based on screenshots and images. This is a dedicated leaderboard that shows which models are the best at agentic coding live sites based on visual inputs.
minor tide
<@&1372208635530448926> - A new model has been added to Text, Vision, Code, Image-to-WebDev, and Document Arena! This model has been added to Battle mode, but is not currently available in Direct or Side-by-Side.
- Claude Opus 4.7
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Video leaderboard and Image-to-Video leaderboard has been updated to include HappyHorse-1.0.
Highlights:
#2 Text-to-Video: Scores 1444
#2 Image-to-Video: Scores 1444
Top 2 for all 3 Video Arena leaderboards
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Code Arena leaderboard, Expert Arena leaderboard & Text Arena leaderboard has been updated to include Claude-opus-4-7 and Claude-opus-4-7-thinking!
Highlights - Opus 4.7 Thinking:
Ranks #1 in Code Arena, +37 points over Opus-4.6, #1 on both React and HTML leaderboards
Ranks #1 in Expert Arena
Ranks #1 in Text Arena, leads across major categories: #1 Coding, #1 Software & IT Services, #1 Writing, Literature, & Language, #1 Life, Physical and Social Sciences, #2 Multi-Turn
Highlights - Opus 4.7:
Ranks # 4 in Expert Arena
Ranks #3 in Text Arena
minor tide
<@&1473460945308221542> <@&1372208524230397962> - In this deep dive, Peter tests Claude Opus 4.7’s capabilities to see how it stacks up against leading models in the Code Arena where agentic web development tasks are evaluated by generating live sites and apps.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Vision Arena leaderboard and the Document Arena leaderboard has been updated to include Opus-4.7 and Opus-4.7-thinking.
Highlights:
Opus 4.7 ranks #1 in Document Arena with a score of 1521, while Thinking ranks #4 with a score of 1508
In Vision 4.7-Thinking ranks #1 while Non-Thinking ranks #3
Vision sub-categories saw the biggest gains over Opus-4.6: #1 Diagram (+20), #1 for Non-Thinking in Homework (+30), and #1 OCR for Non-Thinking (+7)
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard has been updated to include Qwen3.6-plus.
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena, Vision Arena, Code Arena, and Document Arena!
- Kimi K2.6
sand grail
<@&1473460945308221542> See how agentic coding actually works: tool calling, self-correction, multi-turn iteration, and more.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Image Arena leaderboard and the Image Edit leaderboard has been updated to include GPT-Image-2.
Highlights:
#1 Text-to-Image (1512), +242 over #2 (Nano-banana-2 with web-search aka gemini-3.1-flash-image)
#1 Single-Image Edit (1513), +125 over #2 (Nano-banana-pro aka gemini-3-pro-image)
#1 Multi-Image Edit (1464), +90 over #2 (Nano-banana-2)
minor tide
<@&1372208635530448926> - A new model has been added to [Text],(https://arena.ai/), Vision & Code Arena!
- MiMo-V2.5
- Pro versions added to Text and Code
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code, Vision, Document, and Text Arena leaderboards have been updated to include Kimi-K2.6.
Highlights:
#2 open model in Code Arena (#6 overall)
#1 open model in Vision Arena (#15 overall)
#1 open model in Document Arena (#8 overall)
#2 open model in Text Arena (#24 overall)
Stay up to date with our Leaderboard Changelog.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Image Arena leaderboard and the Image Edit leaderboard has been updated to include qwen-image-2.0-pro-2026-04-22.
Highlights:
#9 Text-to-Image
#17 Image Edit (Single Image)
Top 10 in Text-to-Image categories: #6 Portraits, #7 Photorealistic & Cinematic Imagery, #7 Art
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard and Code Arena leaderboard has been updated to include DeepSeek V4.
Highlights:
Code Arena - DeepSeek V4 Pro (thinking): #3 open model (#14 overall)
Text Arena - DeepSeek V4 Pro (thinking): #2 open model (#14 overall), DeepSeek V4 Flash (thinking): #10 open model (#47 overall)
Top 10 Text categories: #1 Medicine & Healthcare (v4 Pro), #8 Legal & Government (v4 Pro), #8 Math (v4 Pro Thinking), #9 Life, Physical, & Social Science (v4 Pro Thinking)
minor tide
<@&1372208635530448926> - A new model has been added to Text, Vision, Search, Document, and Code Arena! Note this is in Battle mode. You will not find this model in Direct and Side by Side modes.
- GPT 5.5
minor tide
<@&1372208635530448926> - We’re updating the rate limits for Image Arena, effective Monday, April 27. This will help us maintain reliability and ensure the service remains sustainable as usage grows. The team is developing a new usage system that'll unify limits, learn more here: #announcements message.
Image Arena Battle Mode will now be limited to 15 generation requests per 24-hour period
The following models will have a 5 generation request per hour limit
- gpt-image-1.5-high-fidelity
- gpt-image-2
- gemini-3.1-flash-image-preview
- gemini-3-pro-image-preview
- chatgpt-image-latest-high-fidelity
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena, Text Arena, Expert Arena, Search Arena, Document Arena, and Vision Arena leaderboards have been updated to include GPT-5.5.
Highlights:
Code Arena: ranked #9, +50 point jump over GPT-5.4
Document Arena: ranked #6
Text Arena: ranked #7, #3 in Math and #8 in Instruction Following categories
Expert Arena: ranked #5
Search Arena: ranked #2
Vision Arena: ranked #5
Stay up to date with our Leaderboard Changelog
!
minor tide
<@&1372208635530448926> - Some users currently have access to our Agent Mode experiment. Agent Mode is a multi-modal chat experience that lets you work across different modalities within a single, unified workflow. Since this is still an experiment,** access is currently limited**. Community feedback plays an important role in how we develop new features, so if you’ve had a chance to try it, we’d love to hear from you.
If apart of the experiment, you will find Agent Mode in the menu drop down where you select Battle, Direct, and Side by Side.
We'll have a few follow-up questions and would really value your input!
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated to include Ernie-5.1.
Highlights:
#13 overall, now the #1 highest-ranked model from a Chinese lab
Categories: #9 Math, #1 Legal & Government, #4 Business, Management & Financial Ops, and #7 Software & IT Services
sand grail
<@&1473460945308221542> GPT Image 2, DeepSeek V4, GPT-5.5, and more — all in 80 seconds
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard has been updated to include hunyuan-hy3-preview.
Highlights:
Top 7 lab for open models in Text Arena
Ranks #80 overall
Priced at $0.29 / $1.17 per 1M tokens
sand grail
<@&1473460945308221542> <@&1372208590248742964> Arena researchers walk through building an autorater from scratch, then get into where it falls apart in practice.
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena, Vision Arena, Document Arena & Code Arena: Front-end.
- Grok 4.3
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Image leaderboard & the Image Edit leaderboard has been updated to include UNI-1.1-Max & UNI-1.1.
Highlights:
Text-to-Image Arena: UNI-1.1-Max #6 overall (1193), UNI-1.1 #7 overall (1190)
Multi-Image Edit Arena: UNI-1.1-Max #7 overall (1315), UNI-1.1 #8 overall (1298)
Single-Image Edit Arena: UNI-1.1-Max #7 overall (1337), UNI-1.1 #11 overall (1310)
minor tide
@everyone - Max, Arena's model router powered by 5M+ community votes, is now multimodal.
Starting today, Max is the default in Direct chat across every modality: search, vision, image generation, image editing, and front-end coding with the same latency-controlled performance as the original router for text.
Learn more about Multimodal Max in this blog post.
minor tide
<@&1372208635530448926> - A new model has been added to Text Arena, Vision Arena, and Document Arena!
- GPT-5.5-Instant
minor tide
<@&1385704933403398274> - We’re hosting a Discord Staff AMA with our Product Lead to answer the most common questions! We’ll be recording the session and sharing it in this server afterward.
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Vision Arena leaderboard has been updated to include Gemma-4.
Highlights:
Gemma-4-31b ranks #2 open (#20 overall)
Gemma-4-26b-a4b ranks #4 open (#26 overall)
sand grail
<@&1473460945308221542> - Gemma 4, UNI-1.1, Grok 4.3, and more — all in 80 seconds
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard, Vision Arena leaderboard, and Document Arena leaderboard has been updated to include gpt-5.5-instant.
Highlights:
Vision Arena: #11 overall, on par with Claude-Sonnet-4.6
Text Arena: #18 overall, Multi-Turn #5
Document Arena: #24, on par with GPT-5.2
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Search Arena leaderboard has been updated to include Ernie-5.1.
Highlights:
Debuts at #4 in Search Arena
Top 3 lab in Search
The only Chinese model in the top 10 overall
minor tide
<@&1372208524230397962> <@&1372208635530448926> - Introducing 7 new leaderboard views for frontend output in Code Arena. Aggregate leaderboards don’t tell the full story. "Best frontend coding model" depends on what you're building, so we built leaderboards that show exactly that.
Here are the new major frontend web development task categories:
Brand & Marketing
Reference-Based Design
Data & Analytics
Consumer Product
Gaming
Simulations
Content Creation Tools
minor tide
<@&1372208635530448926> - We’re currently reworking Arena’s usage system into a single, unified one. Instead of separate hidden limits across different features and models, the new system is designed to give users a daily balance of credits that can be used however they want across Arena. The goal is to provide much more visibility, flexibility, and control over usage.
As part of the rollout for the new usage system, we’ve recently added daily limits for all Arena users. You may notice a new error message letting you know when you’ve hit your daily usage limit, along with info on when it resets. Your usage will automatically reset after 24 hours.
These limits are being put in place ahead of the full rollout to help support upcoming features like Agent Mode and improve overall system stability while we continue building out the remaining parts of the new usage system. The daily usage limit is one part of the rollout while we continue building the rest of the system. Additional features, including clearer visibility into daily usage spending and other usage management tools, are still in development.
If you run into any issues with the system, please report them in this thread: #1503450500614455337 message
sand grail
<@&1473460945308221542> <@&1372208635530448926> Peter tests the top four open source models out of China — GLM 5.1, Kimi K2.6, MiMo 2.5 Pro, and DeepSeek V4 Pro
minor tide
<@&1372208635530448926> - Now that Agent Mode has rolled out to more users, we’re really hungry for feedback.
If you’ve tried out the new mode, we’d love to hop on a casual voice call with you sometime in the next couple of days. If you’re interested, just ping me in the #general.
If you’d prefer to share feedback async instead, we also have a few follow-up questions here.
Let me know if you have any questions! Really appreciate everyone taking the time to try it out.
sand grail
<@&1473460945308221542> <@&1372208590248742964> Arena researchers dive deep into the data pipeline behind Arena's tagging system
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard and the Vision Arena leaderboards has been updated to include Qwen3.7.
Highlights:
Qwen3.7 Max Preview ranks #13 overall in Text
Qwen3.7 Max Preview ranks #16 overall in Vision
Standout categories include #7 Math, #9 Expert, #9 Software & IT, and #10 Coding
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text Arena leaderboard and the Code Arena leaderboard has been updated to include Gemini 3.5 Flash.
Highlights:
Ranks #9 overall for both Text and Code: Frontend
Sub-category highlights: #7 Content Creation Tools, #8 Gaming, #8 Consumer Product, #9 Data & Analytics, and #10 Reference-Based Design
minor tide
<@&1372208635530448926> - We are changing some of the models available in Direct and Side-by-Side modes for Image Arena. These changes are part of our efforts to ensure that we can continue offering access to AI models while keeping the platform running reliably.
The following models are being removed from Direct and Side-by-Side on <t:1779469200:F>
- gemini 3 pro image preview 2k
- gemini-3.1-flash-image-preview
- gpt image 1.5 high fidelity
- gpt image 2 medium
Updates like this help us maintain availability for everyone over the long term, and we intend to bring these models back in a way that’s more sustainable when possible.
sand grail
<@&1473460945308221542> Gemini 3.5 Flash is doing more than you ask
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Text-to-Image Arena leaderboard has been updated to include HiDream-O1-Image.
Highlights:
Ranks #27 overall
Ranks #4 open source for Text-to-Image Arena
minor tide
<@&1372208524230397962> <@&1372208635530448926> - The Code Arena leaderboard has been updated to include Qwen3.7 Max (20260517).
Highlights:
#4 in Code Arena: Frontend
Scoring 1541
