No capacity available

[kplgr]

I have moved this to an issue as there is zero response from the developers

Issue link: #20008

I keep getting these errors, eg
✕ [API Error: No capacity available for model gemini-3-flash-preview on the server]

This has been going on ALL DAY today. For every model.

My question is , whose servers are out of capacity exactly? Definitely not google's...

By the way, I have used OAuth with a Google AI Pro-enabled account.

[bronyhomie365]

Same here - any ETA on the fix please?

[gustavofrdev]

same here. I Pay for something that is not avaible.

[Canahmetozguven]

same here

[jkuk06]

Same here. With enterprise account.

[yurifilgueira]

Same here.

[mrowlinson]

Yet another. I'm assuming this is somehow related to the launch of 3.1 Pro but it's still ridiculous. I pay for this product and Google is routing me to years old models.

[salmanbappi]

Same here

[wpted]

same here

[KyunghyunLee]

Same here. My account has plenty of usage remaining, but I am getting this message continuously.

[kiyoaki]

Experiencing the same issue. gemini-3-flash-preview returns 429 MODEL_CAPACITY_EXHAUSTED intermittently.

Environment:

• Gemini CLI v0.29.5, Windows 11
• Google AI Ultra for Business (OAuth)

Key observation:
When using -m gemini-3.1-pro-preview, the main model prompt succeeds, but internal tool calls fail with 429. The error message references gemini-3-flash-preview, suggesting that tool-call requests are routed to flash regardless of the model I specified. This makes the entire session unusable even when only flash capacity is exhausted on the server side.

/stats output at time of error:

• gemini-3-flash-preview: 75.4% remaining
• gemini-3.1-pro-preview: 93.2% remaining

So this appears to be a server-side capacity issue, not a per-user quota limit.

It would be helpful if the CLI could fall back to an alternative model for internal tool calls when the primary one hits capacity limits.

[BrianInAz]

It's the middle of the night here and still getting : No capacity available for model gemini-3-pro-preview on the server?!?!

[zhangtaile]

same here

[PS-Wizard]

+1 no capacity available for gemini-3-pro-preview on the server

[partyalarm]

After a while, it changed to the following error.
#17643
#18206

[manas-17045]

+1 No capacity available even for gemini-3-flash-preview

[Chrys4lisfag]

ULTRA SUBSCRIBER BTW

[arlucio]

2 weeks already as an Ultra subscriber and I cant use any model via gemini-cli or get any response, I live in japan and never felt so disrespected by any company

Not sure if I can join a law process in the US for this but I woudl be ok to spend money to support some law measure against google

[dhaern]

Dont pay Ultra suscription, this is an absolute scam having chatgpt suscriptions cheaper and models 100% working with better performance and you can use them with 3rd party software.

[arlucio]

Unfortunately my subscription was in auto renewal an it got renewed, I will not be renewing from now on

[YoshiaKefasu]

Anthropic and Google are tricking users into building their AI infrastructure, only to exploit even more people, lol. OpenAI was the first, and others jumped on the hype that came after. The AI model itself was solid, but the company’s motives were disappointing.

[applepig]

Suggestion: "gemini-3.1" to rename as "gemini-429"

[alphaply]

now it is gemini-403

[zhangtaile]

any update? the issue is still here

[JimChengLin]

just give up gemini cli, that sucks

[dhaern]

just give up gemini cli, that sucks

The problem is when we use OUR suscription and Oauth in 3rd party software, these Google rats unfair ban us.

[kyleboddy]

ultra subscriber here and still these problems. joke model provider while they tweet about how good their stuff is

HTTP 429 "No capacity available for model gemini-3-flash-preview"

[ihr-it-projekt]

Same here...

[ihr-it-projekt]

relog into anti gravity helps a bit

[HA-HUB-I]

same here

[Kovinda]

same here... No capacity available for model gemini-3.1-pro-preview on the server

[QingyeSC]

same here... No capacity available for model gemini-3.1-pro-preview on the server

[michail-p]

same here

[Luckyfive]

I've been experiencing this more and more. It randomly goes away sometimes

[hppsc1215]

same issue

[guilhermezaparoli]

same here

[xikass]

same here

[autom8edIT]

By the way, I have used OAuth with a Google AI Pro-enabled account. <-- Make a project / API key and remove the OAuth since the cli don't understand that you have that sub.
The /auth is just a decoy, it only sign you in, it doesn't check if you've got a sub or not

https://console.cloud.google.com/home/dashboard <-- Here you can create a project and API key
The run these two commands in an elevated powershell terminal.
gcloud config set project project-XYZ123
gcloud auth application-default set-quota-project project-XYZ123

The Cloud will still not work most of the time anyone in your timezone is awake but now you've at least got a small chance of it working, sometimes.basically not work until they make a serious backend, as long as the cli rely on node.js there's 0% chance of it working.

[40960]

{
  "error": {
    "code": 503,
    "message": "No capacity available for model gemini-3.1-flash-image on the server",
    "status": "UNAVAILABLE",
    "reason": "MODEL_CAPACITY_EXHAUSTED"
  }
}

I feel like Google is scamming me out of my money.