I am writing to report severe, recurring quota throttling, slowness on my account
to anything Gemini related. despite paying for the highest tier. Since September 11,
my workflows have been crippled by constant HTTP 429 RESOURCE_EXHAUSTED
errors while actual usage is a tiny fraction of my advertised limits.
Instead of the throughput promised for pro plans, my requests hit backend
locks across models, including Gemini 3.8 Flash and Antigravity tooling.
It doesn’t fail outright, sometimes stalls in prolonged exponential backoff
loops, turning a normal 4 second tasks into 40+ second freezes. Even
minimal agentic tool calls trigger immediate locks across sessions.
I’m paying for maximum capacity only to encounter artificial restrictions, it
defeats the entire purpose of this professional subscription. I am asking the
engineering team to investigate backend synchronization for my account,
clear the false exhaustion flags, and restore the actual quota limits tied to
my paid plan.
The HTTP 429 Too Many Requests (RESOURCE_EXHAUSTED) error typically indicates that your request volume has exceeded the per-minute rate limits.
Ensure you have updated to the latest available version of Antigravity (v2.13.0) or the Antigravity IDE (v2.5.5)
If the issue still persists after upgrade, open the Command Palette (Ctrl+Shift+P) in Antigravity and run the command: Antigravity: Reset onboarding to force a state refresh
If the issue persists after the reset steps above, please share the latest trajectory ID.
I also want to confirm: Do these errors occur during web searches? Web searches use a special agent to summarize and it’s often unavailable. Instruct your agent to search with &udm=18, curl or DuckDuckGo if you’re doing web stuff.
Hello @Lea. web searches seem fine so far. The major issue is recent slowness and it seems consistent across Sao Paulo, Brazil and Lynchburg, America/New_York regions.
This is what i did to resolve my own 429 errors, i wrote a big int patcher for antigravity. pretty sure the issue is there is some mangled context which is too large a packet to send successfully to google ai servers. this results in immediate denial of service request. basically, the “resource” that is “exhausted” is the local context. there is a release for linux and apple and windows. there is also a cleanup feature to clean cache and cookies for antigravity, so that when you do reinstalls or updates, you have the best chances of reducing upgrade errors.
i should mention as well, this issue has persisted for the entire year since gemini 3.1 - generally the primary central issue, is multiple google accounts on the computer and the cookies and authentication get confused with multiple local google accounts. this compounds even moreso when there are workspace accounts involved. i have a massive post on reddit about this detailing my entire investigation. i began working on a software based solution to prevent account identity mismatch, but it is incredibly complex and incomplete due to the advances of internet security, by the time my own finalized perfected solution is completed, google might fix it on their own. but regardless, the reddit article is a great read to help understand how this came to be and what the process looks like as a user finding a solution.