Persistent 429 RATE_LIMIT_EXCEEDED when using generate_image with gemini-3.1-flash-image in Google Antigravity 2.11.0

Google Antigravity
1

Summary

I am consistently unable to use the generate_image tool in Google Antigravity.

Every request to gemini-3.1-flash-image through cloudcode-pa.googleapis.com returns:

HTTP 429 RESOURCE_EXHAUSTED
reason: RATE_LIMIT_EXCEEDED

The unusual part is that the returned retryDelay is very short, but waiting significantly longer does not restore access.

I am using Google AI Pro, and Antigravity correctly recognizes my plan.

Environment

  • Antigravity version: 2.11.0

  • Release channel: Stable

  • Google plan: Google AI Pro

  • Authentication: Google OAuth

  • Tool: generate_image

  • Model: gemini-3.1-flash-image

  • Backend: cloudcode-pa.googleapis.com

  • OS: Windows

Reproduction

The issue can be reproduced with a single image-generation request.

For example:

Generate a simple red apple on a white background.

The request is made only once.

There is:

  • no concurrency

  • no parallel agent

  • no automatic retry

  • no repeated tool call

The request still returns HTTP 429.

First observed response

{
  "error": {
    "code": 429,
    "message": "You have exhausted your capacity on this model. Your quota will reset after 3s.",
    "status": "RESOURCE_EXHAUSTED",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.ErrorInfo",
        "reason": "RATE_LIMIT_EXCEEDED",
        "domain": "cloudcode-pa.googleapis.com",
        "metadata": {
          "uiMessage": "true",
          "model": "gemini-3.1-flash-image",
          "quotaResetDelay": "3.643589217s",
          "quotaResetTimeStamp": "2026-08-28T08:16:14Z"
        }
      },
      {
        "@type": "type.googleapis.com/google.rpc.RetryInfo",
        "retryDelay": "3.643589217s"
      }
    ]
  }
}

Waiting test

I then stopped making image-generation requests and waited approximately 15 minutes before trying again.

The next single request still returned:

429 RESOURCE_EXHAUSTED
reason: RATE_LIMIT_EXCEEDED

with another short retryDelay.

Strict retry-delay test

I also performed a controlled test specifically following the server-provided retry delay.

Request 1

Timestamp:

16:29:13

Response:

"reason": "RATE_LIMIT_EXCEEDED",
"quotaResetDelay": "177.423646ms",
"retryDelay": "0.177423646s"

I then deliberately waited approximately 12 seconds, which is substantially longer than the reported 177 ms retry delay.

Request 2

Timestamp:

16:29:25

The second request was again a single, non-concurrent request.

It returned:

{
  "error": {
    "code": 429,
    "message": "You have exhausted your capacity on this model. Your quota will reset after 0s.",
    "status": "RESOURCE_EXHAUSTED",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.ErrorInfo",
        "reason": "RATE_LIMIT_EXCEEDED",
        "domain": "cloudcode-pa.googleapis.com",
        "metadata": {
          "quotaResetTimeStamp": "2026-08-28T08:29:26Z",
          "uiMessage": "true",
          "model": "gemini-3.1-flash-image",
          "quotaResetDelay": "167.621173ms"
        }
      },
      {
        "@type": "type.googleapis.com/google.rpc.RetryInfo",
        "retryDelay": "0.167621173s"
      }
    ]
  }
}

Therefore, even after waiting approximately 12 seconds, the request was still rejected with the same RATE_LIMIT_EXCEEDED error.

Additional observations

The problem appears to be isolated to image generation.

The main text-generation model in the same Antigravity installation works normally, with no 429 errors.

Antigravity also correctly identifies my account as:

Google AI Pro

The local installation is also running the current Stable version available to me:

Antigravity 2.11.0

What makes this unusual

The server repeatedly reports a very short retry delay:

~0.17s
~3.6s

However, waiting substantially longer does not make the next request succeed.

This has happened across multiple independent tests:

Single request                  → 429
Wait several seconds            → 429
Wait ~15 minutes                → 429
Single request, no retry        → 429
Wait retryDelay + 5 seconds     → 429

All failures consistently identify:

reason: RATE_LIMIT_EXCEEDED
model: gemini-3.1-flash-image
domain: cloudcode-pa.googleapis.com

Questions

Could someone confirm whether this is expected behavior for gemini-3.1-flash-image under Google AI Pro in Antigravity?

Specifically:

  1. Does retryDelay in this response only indicate a minimum retry interval rather than actual model availability?

  2. Is there an additional account-level, subscription-level, or model-level rate limit that is not exposed in this error response?

  3. Is gemini-3.1-flash-image currently experiencing a known capacity/rate-limiting issue in Antigravity?

  4. Is there any way for users to check the actual quota/capacity status for this model?

  5. Is this a known issue affecting Google AI Pro users?

I would appreciate confirmation of whether this is an expected Pro-tier limitation or a backend/Antigravity issue.

I can provide additional raw error responses and diagnostic information if needed.