Persistent 503 MODEL_CAPACITY_EXHAUSTED on claude-opus-4-6-thinking

Model: claude-opus-4-6-thinking
Plan: Antigravity pro

Issue:
Every request to claude-opus-4-6-thinking fails with a 503 capacity error from the Cloud AI Companion backend. This is blocking my ability to use the model.

Error details:
Trajectory ID: a645ce8f-2fe9-4a24-ba8c******
TraceID: 0x86aa2c4******
HTTP Status: 503 UNAVAILABLE
Domain: cloudcode-pa.googleapis.com
Reason: MODEL_CAPACITY_EXHAUSTED
Message: “No capacity available for model claude-opus-4-6-thinking on the server”
Date: Wed, 09 Sep 2026 14:41:09 GMT

Steps to reproduce:

  1. Open Antigravity
  2. Select claude-opus-4-6-thinking as the model
  3. Send any request
  4. Request fails immediately with the above 503 error

What I’ve tried:

  • Retrying multiple times
  • Switching to claude-sonnet-4-6-thinking
  • Waiting several hours

Impact: Blocking my daily coding workflow and website coding

Could the team confirm if this is a known capacity issue and provide an ETA for resolution?

Hi @pumkin,

Welcome to the Forum,

A 503 Unavailable (MODEL_CAPACITY_EXHAUSTED) error indicates that the server is temporarily overloaded or experiencing a capacity constraint. This occasionally occurs during peak usage hours.

We have verified and it appears to be serving requests normally at the moment. Could you please check and confirm if you are still experiencing this issue?

If you encounter this error again, the best immediate workaround is to switch to Gemini 3.1 Pro in the model selector. Its serving capacity is highly stable and should allow you to work without interruption.

Hi, we expirience this error for a week, we started testing our gemini integration and it keeps happening.
Reproducing currently, model gemini-flash-latest