Issue: Google AI Studio Build repeatedly returns “Quota exceeded. Please try again later.”
Tier: Paid Tier 1
Model: Gemini 3.8 Flash
Visible rate limits: RPM 19/1000, TPM 1.84M/2M, RPD 434/10K
Antigravity: 0/30 RPM, 0/200K TPM, 0/1K RPD
Observed behavior: Build agent runs for several minutes, makes/builds changes successfully, then terminates with quota exceeded.
Typical run duration: approximately 600–700 seconds.
Billing: Active and below spend limits.
Request: Please check the AI Studio Build/Agent entitlement, execution quota, or backend/provider-routing state for this project because the visible Gemini API quota is not exhausted.
Hi @Nerissa_Organo,
Seems like your token usage was almost full when you reported this. Please note that the rate limits shown on AIS page have a slight lag, so its likely that you had exceeded the TPM code before it was visible on the rate limits page.
Same here. I traced it in my project: RPM and RPD were almost empty, but TPM was at 2.44M / 2M. The Build agent resends the whole conversation and code context on every turn, so a long session blows the tokens-per-minute limit. Full breakdown and workarounds that worked for me: