Persistent HTTP 503 Error on Gemini 3.8 Flash Medium – Account-Specific Issue

Hello Google Support,

I am experiencing persistent HTTP 503 errors when using the gemini-3.8-flash-medium model through Antigravity.

After further testing, I discovered an important detail:

On the same device, using the same network connection and the same Antigravity application, I tested the model with a different Google account. The new account works normally without any errors. However, my original Google account continues to experience intermittent HTTP 503 errors.

This suggests that the issue may be specific to my original Google account rather than a device, network, or general service availability issue.

The error details are as follows:

  • HTTP Status: 503 Service Unavailable

  • Error Reason: MODEL_CAPACITY_EXHAUSTED

  • Error Message: No capacity available for model gemini-3.8-flash-medium on the server

  • Error Number: 2010

  • Model: gemini-3.8-flash-medium

  • Trajectory ID: `251b34d3-6a8e-44f1-b64b-1866f1f51..

  • Trace ID: `0x8368ecb147176..

Could you please investigate whether my original Google account has any account-specific restrictions, resource allocation issues, quota limitations, or other backend problems?

Since the same model works normally with another account under identical conditions, I would appreciate a specific investigation into the status of my original account.

Please also advise whether any action is required to restore normal service.

Thank you for your assistance.

Hi @Dian,

Thanks for sharing the full error details and testing with a second account—that’s really helpful context. While we can’t directly inspect individual Google account records from the forum, an HTTP 503 Service Unavailable error (“No capacity available”) is a server-side model capacity issue rather than a problem with your device, network, or local Antigravity installation. You don’t need to reinstall the app or clear local data for this. In the meantime, you can try switching Gemini 3.8 Flash to a different thinking level (Low or High) in the model selector under the prompt box, or switch to another model available on your plan (such as Gemini 4 Argon on a paid Google AI Pro or Ultra plan, or Claude Sonnet 4.6 (thinking) / GPT-OSS-120b on Free or Pro). You can also check Settings in Antigravity to verify your plan tier and baseline quota usage.

Let me know is the issue resolves or the issue still persists!

Hi, thank you for the suggestions. I have thoroughly tested all the steps, but the issue persists with very specific behavior:

  1. Testing Thinking Levels (Low / Medium / High):
    • On my primary account (Pro tier), switching Gemini 3.8 Flash across Low, Medium, and High thinking levels still results in the exact same HTTP 503 (MODEL_CAPACITY_EXHAUSTED).
  2. Other Models on Primary Account:
    • On this same primary account, every other model (including Claude, GPT, and other available versions) works flawlessly with sub-second response times. The issue is strictly isolated to the Gemini 3.8 family.
  3. Multi-Account Comparison:
    • When switching to my wife’s account (a Free tier account) on the exact same machine, network, and Antigravity installation, Gemini 3.8 Flash works normally without any 503 errors.
    • However, switching back to my primary Pro account immediately triggers the 503 error on 3.8 again.

This strongly indicates that my primary Pro account is pinned to an exhausted/degraded backend tenant cell or a locked quota slice for Gemini 3.8, while Free accounts or other models are routed to functional pools.

Could you please escalate this to have my primary account’s backend quota mapping or server-side routing slot for Gemini 3.8 refreshed or migrated to an active pool?

Thank you!

Hi @Ambati_Rajendra ,
I noticed multiple users on the forum reporting identical issues. Could you please confirm if this has been officially logged as a known issue, and whether the root cause has been identified by the backend team?

Hi Sorry to bother you again. As of today, the issue remains unresolved.

Could you please provide an update on the progress or let me know the current status?

Thank you