#Gemini Exp 1114

114 messages · Page 1 of 1 (latest)

rugged totem
#

Gemini Exp 1114

jade dagger
#

Gemini Exp 1114

past fractal
#

Geminu Exp 1114

spark apex
#

Wut wut

tawny mural
spark apex
#

Yeah but it's not available via an API yet xd

twilit steeple
wanton surge
#

supposedly it's supposed to work already

#

(gemini api, not vertex api)

spark apex
twilit steeple
#

It's still AI Studio only - if you try and call it via the API you just get a 404. It'll also probably be highly rate limited like the previous experimental models

rugged gazelle
#

there has to be suspicion that this model is gemini-2.0-pro-0111 experimental

rugged totem
#

Its now available

hollow hamlet
#

How can you tell? The page was the same before too

smoky cliff
#

The label has changed for AI studio provider since yesterday but that's probably nothing

#

But the page linked by @rugged totem is exactly the same as it was yesterday yes.

teal peak
celest bloom
#

@spark apex plz add now?

spark apex
#

lmk if yall see issue xd

#

RIP already 429s

celest bloom
#

U guys are definitely like a big customer now right

smoky cliff
#

sounds logical but it's google y'know

#

AND it's an experimental model

spark apex
celest furnace
#

we dont even know the pricing

#

and exp have horrible limits

dusk shuttle
#

Context for it is 1m? Not 32k like in Ai studio?

low pasture
celest furnace
#

need to use for batching

low pasture
celest furnace
twilit steeple
#

it shares the same rate limits as the 1.5 pro experimental model, for a cool 2 rpm, 1mtok/m, 50 rpd

twilit sequoia
spark apex
twilit sequoia
#

i'm literally copy and pasting the wikimedia image demo code

#

it works as long as the model isn't this one

spark apex
#

Rip

twilit sequoia
# 
import requests
import json

response = requests.post(
  url="https://openrouter.ai/api/v1/chat/completions",
  headers={
    "Authorization": f"Bearer {OPENROUTER_API_KEY}",
    "HTTP-Referer": f"{YOUR_SITE_URL}", # Optional, for including your app on openrouter.ai rankings.
    "X-Title": f"{YOUR_APP_NAME}", # Optional. Shows in rankings on openrouter.ai.
  },
  data=json.dumps({
    "model": "google/gemini-exp-1114", # Optional
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "type": "text",
            "text": "What's in this image?"
          },
          {
            "type": "image_url",
            "image_url": {
              "url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg"
            }
          }
        ]
      }
    ]
    
  })
)
spark apex
#

It worked in aistudio XD....

twilit sequoia
#

not sure what to say...

spark apex
#

Hmmm that's weird

#

Can you try a png img?

twilit sequoia
#

{'error': {'message': 'Provider returned error', 'code': 429, 'metadata': {'raw': '{\n "error": {\n "code": 429,\n "message": "Resource has been exhausted (e.g. check quota).",\n "status": "RESOURCE_EXHAUSTED"\n }\n}\n', 'provider_name': 'Google AI Studio'}}}

#

did you make google mad 😉 ?

#

too many people going nuts today?

low pasture
#

Wait, try again.

twilit sequoia
#

still "exhausted"

low pasture
#

You will see a ton of these errors for all experimental models from Google, get used to it.

#

With "wait" I mean a minute, at least.

twilit sequoia
#

is that just via openrouter or with vertext/etc as well?

low pasture
smoky cliff
#

a jpeg worked for me (But i sent it in base64 and not using an url)

stuck parrot
#

smh why can't Google have actual pay plans for experimental models?

low pasture
stuck parrot
#

not enough hardware dedicated to it? (as in they literally wouldn't be able to serve more)

#

ok, makes sense if so

#

that or to prevent an oopsie if there's some weirdness, so companies don't go using it at large scale

dull nova
low pasture
#

Complaints about experimental/preview models which are free of charge are pointless. They either work or they don't.

formal lion
#

Just use Google AI Studio directly, it's still free

celest bloom
#

@spark apex Could u please ask google for like a rise in rate limit, really wanna use it on OpenRouter

spark apex
#

Tho I doubt we can get bump on experimental models

#

But we will have something for this soon -- i.e rolling out of BYOK

celest bloom
#

Why does Gemini even has these experiments

#

they should just push to prod

#

Gemini 114 is definetly good enough as a prod model

spark apex
#

likely capacity ramping up

#

you can't just summon TPUs

celest bloom
teal peak
#

1121 is even better tbh

celest bloom
spark apex
#

Just pushed all the models up

#

FYI moved them all to a free variants

celest bloom
#

@spark apex

spark apex
celest bloom
#

@spark apex Can I get the option to slap in my own API key for these models

#

like my own google models

#

I see you guys have intergrations

spark apex
celest bloom
#

I kinda wanna use it today with the OpenRouter API

#

too lazy to set up gemini api

spark apex
#

cc @lavish nova

#

We removed it a while back fam

#

#announcements message

#

Also that's off-topic for this thread

#

(this is for gemini 1114)

#

Not at this time!

low pasture
#

Wrong channel, you have been told this before. Use #arc-requests or #arc-feedback

molten skiff
#

why do a ton of the gemini models throw the 429 error?

lavish nova
#

Which ones? if they're called "experimental" in the name, that's the reason why

echo creek
#

how i can get a free api for this model ?

lavish nova
wanton surge
#

and use gemini api

shadow storm
low pasture
#

Google's experimental models will always be next to unusable through OpenRouter

wanton surge
#

just use aistudio

shy zinc
#

Does anyone know if google/gemini-pro-1.5 is updated with new versions of the experiment models or if it still has an old version?

wanton surge
#

I think it's the old version

#

@shy zinc if you need the experimental model you should get an api key from aistudio.google.com and use the gemini api; the rate limit is too low for you to use it well on openrouter