Hi all — I’m running into a persistent MODEL_CAPACITY_EXHAUSTED (503) error, but only on my account in my workspace.
Context:
I’m on Ultra
A colleague in the same workspace, same domain/email setup, also on Ultra → no issue on their side
This happens consistently for me on claude-opus-4-6-thinkingclaude-sonnet-4-6-thinking
Even on the first prompt of the day, sometimes after just a few seconds of execution.
So far it looks account-specific rather than workspace-wide.
Has anyone else seen this kind of discrepancy between users in the same workspace? Any ideas what might cause capacity errors to affect only one account?
I am a Business AI Ultra user. If you are seeing the “Error: HTTP 503 Service Unavailable” message, I unfortunately have bad news for you. Support has no idea what is causing this, or they just don’t want to fix it. This has been going on for months, and the only real fix is to unsubscribe. Have fun.
We’re finding new ways to share more capacity with the Antigravity community. During windows where we have unused capacity, currently in the late afternoon PST, you’ll notice that your baseline quota stretches further. You’ll get more requests and less downtime within your existing plan.
As global demand shifts, we’ll continue to adjust these windows to pass any available capacity back to you. We’re committed to expanding access and ensuring every developer has the capacity they need to keep building.
so i solved mine with this BigInt patcher, it helped a lot but its just one thing and theres a LOT more going on with antigravity than 1 thing to fix to even address the full root cause of the issue.
note : since its my own ware i’m glad to provide support, run the patcher for “easy” patches, if that resolves you’re good until the next update. if it doesn’t resolve, try “risky” patches. any problems, theres the cleanup utility in there too to clear cache for truely clean reinstalls.
I’m experiencing the same issue with AI Ultra. It occurs within a single workspace following an extended workflow. It appears the context window has grown too large for the model to process the instructions effectively, resulting in a 503 error.