AI Studio Build Mode repeatedly returns "Quota exceeded" on Paid Tier 2

Google AI Studio
1

We are using Google AI Studio Build Mode on a Paid Tier 2 Gemini API project.

The Gemini API Rate Limit dashboard clearly shows that Tier 2 is active. Our usage is also well below the listed Tier 2 limits for Gemini 3.5 Flash and Gemini 3.7 Flash.

However, Build Mode repeatedly fails with:

“Quota exceeded. Please try again later.”

and sometimes:

“Encountered retriable error from model provider: Agent execution terminated due to error.”

This has continued since yesterday, even after waiting overnight.

We also switched from Gemini 3.7 Flash to Gemini 3.5 Flash, but the same problem continues.

The agent sometimes works for 1–3 minutes, successfully edits project files, and then terminates with the quota error. So the issue is intermittent, but it prevents us from reliably using Build Mode.

Current account/project state:

  • Paid Tier 2 is active
  • Prepaid balance is positive
  • Auto-reload is enabled
  • Gemini 3.5 Flash usage is far below the visible Tier 2 limits
  • Gemini 3.7 Flash is also below the daily request limit
  • The issue occurs specifically during AI Studio Build/agent runs

Could the AI Studio team please confirm whether Build Mode has a separate agent quota, project-level quota, backend capacity limit, or other restriction that is not shown in the Gemini API Rate Limit dashboard?

I can provide screenshots of the Tier 2 rate-limit dashboard and multiple Build Mode runs ending with “Quota exceeded.”

tier limit
tier limit1845×599 49.7 KB

error2
error21812×692 51 KB

error chat
error chat647×1525 63 KB

2

Hello @Zaga_Medya ,

I have DM`ed you for more information .

3

I am also getting the same error. I have paid Gemin i API as well . How to fix it ?

4

Hello @NILADRI_BIHARI_NAYAK ,

Can you please share your App ID, tier, and billing ID via DM?

Which tier are you on—Free or Tier 1, 2 or 3?

5

Dear Mahesh,

My App ID is 49c7ef45-a412-4259-afda-33a24a5094c..

  • Billing ID: 01B35D-FEFA28-25FD..

  • Tier: Paid 1 (Tier 1)

6

Google AI Studio Build repeatedly reports “You are out of generations for this project” even after switching to newly created paid Tier 1 projects and new Gemini API keys.

I tested multiple projects including realestate-master-dev, realestate-master-dev2, and realestate-master-dev3, all with billing enabled and valid Gemini API keys.

The error occurs immediately in Build mode, while Firebase/Firestore, GitHub synchronization, billing, and the application itself are working normally.

This appears to be an AI Studio Build generation/agent quota or backend entitlement issue rather than a Gemini API key or billing issue.

Please check the Build generation entitlement/quota state for my account/projects.

7

realestate-master-dev3 is correctly connected with billing enabled and a valid Gemini API key, but AI Studio Build still immediately reports that the project generation limit has been reached.

8

Hello @gaea67 ,

I have DM`ed you for more details .

9

i am also getting the same error

I’m experiencing the same issue on a paid Tier 1 account. AI Studio Build starts normally, runs for several minutes, partially modifies the project, and then terminates with “Quota exceeded. Please try again later.”

My Gemini API Rate Limit dashboard shows available quota and I am not near the listed RPM/RPD limits. I also created a completely new AI Studio project and the same issue persists.

One failed Build run using Gemini 3.8 Flash ran for 661 seconds before returning the quota error. The dev server itself starts normally and Vite reports connected, but the agent terminates before completing the application.

Tier: Tier 1
Model: Gemini 3.8 Flash
App ID: 0d42b7ac-f6eb-4b87-8fa9-9dd62ed934f8

Could you please check the Build generation entitlement/quota state on my account? I can provide my Billing ID and project details privately via DM.

10

I am experiencing the same issue in Google AI Studio Build.

Model: Gemini 3.8 Flash
Environment: AI Studio Build on desktop web

The Build agent operates normally for several minutes — reading files, editing code, running terminal commands, and tests — and then begins repeatedly returning:

“Encountered retryable error from model provider: Agent execution terminated due to error.”

After several retries it reports:

“Quota exceeded. Please try again later.”

Selecting “Finish what you were doing” sometimes allows the agent to resume and continue successfully for several more minutes before the same provider termination occurs again.

One affected run operated for approximately 624 seconds before failure; after resuming, another operated for approximately 289 seconds and made additional repository edits before failing again.

11

Paid Tier 2. AI Studio Build using Gemini 3.8 Flash repeatedly terminates with “Quota exceeded. Please try again later.”

Attempt 1: 548 seconds
Attempt 2: 558 seconds

Gemini API dashboard is well below Tier 2 limits: Gemini 3.8 Flash peak 26/2K RPM, 1.78M/3M TPM and 888/100K RPD. Prepaid balance is positive.

The near-identical ~9-minute termination time suggests a Build/agent execution quota or backend limit rather than the published Gemini API RPM/TPM/RPD limits.

Please check the Build/Antigravity agent entitlement or execution quota for my App ID/project.

12

Hello @JP_Le_Riche , @Praj_D ,

I have DM`ed you for more detail .

13

I am experiencing what appears to be the same problem on a Paid Tier 2 project.

AI Studio Build repeatedly terminates with:

“Quota exceeded. Please try again later.”

My Gemini API Rate Limit dashboard confirms that I am on Tier 2.

Initially I suspected that the problem might be caused by my existing application, because that application connects to a large database and has a long AI Studio development history.

However, I performed a controlled test by creating a completely new AI Studio Build application with this prompt:

“Create a simple webpage that displays ‘Hello World’ and has a button that changes the text to ‘Hello Art.’”

The new application contained no large database and no previous development history.

The new Build ran using Gemini 3.8 Flash for 593 seconds and then terminated with exactly the same error:

“Quota exceeded. Please try again later.”

Therefore, the problem can be reproduced in a brand-new trivial application and does not appear to be caused by the size, database, or conversation history of my original application.

I can provide my App ID, Google Cloud Project ID, Billing ID, screenshots of the Tier 2 rate-limit dashboard, and screenshots of the failed Build run privately.

Could someone from the AI Studio team please check the Build generation entitlement/quota state for my account/project and determine whether this is a Build/agent quota or backend issue that is not reflected in the Gemini API Rate Limit dashboard?

14

I am experiencing the same issue on Paid Tier 1.

AI Studio Build mode has been showing “Quota exceeded. Please try again later.” for more than 60 hours.

I still have a positive prepaid Gemini API balance (£9.22), and my app preview continues to work. The issue appears to be specifically with the AI Studio Build coding agent.

The last Build run completed its code edits and build step before the quota error appeared. I have avoided repeatedly retrying, but the Build agent is still unavailable.

Could the Google AI Studio team please check my account/project as well?

I can provide my AI Studio App ID, Google Cloud Project ID, billing details, and screenshots privately by DM if required.

Thank you.

15

Hello @anushree_patil , @Art_Kohn

I have DM`ed you for more detail .

16

I’m experiencing the same issue on the Free tier.

AI Studio Build mode keeps returning “Quota exceeded. Please try again later.” repeatedly, across multiple models (tried Gemini 3.8 Flash and Gemini 3.7 Flash) and even after starting a brand new chat/session.

However, when I check the Rate Limits page for my project, RPD shows 0/20 for both models — so there’s clearly no real quota being consumed. This has been happening consistently, not just once.

Project ID: inoutfarma-rececoes-497613
Tier: Free

I’ve also tried: logging out/in, refreshing, switching models, and starting new chats — none of it resolved the issue. Checked status.cloud.google.com and there are no broad incidents reported.

Happy to share screenshots or more details if useful. Hoping this can be looked into given how many of us seem to be hitting it.

17

Additional context: I was making changes to this same app without any issues just a couple of days ago. The “Quota exceeded” errors only started yesterday, with no changes on my end — same project, same models, same usage pattern as before. This suggests something changed on the backend recently rather than anything specific to my account/project.

18

Additional technical detail that might help diagnose this: I had the Build agent investigate the crash directly. It confirmed the exported APK’s classes8.dex is flagged as invalid by dexdump, with errors like:

“Interface virtual method … offsetSize-PENXr5M is not public”

This happens on multiple interfaces in that dex file (androidx.compose.ui.graphics.drawscope.DrawScope, coil.size.ViewSizeResolver, protobuf’s MessageLite), which is illegal bytecode — interface methods must always be public. This looks like a D8/R8 compiler bug that produced corrupted bytecode during the build, not an issue with app code or my prompts. The ART runtime then throws IncompatibleClassChangeError when loading these malformed interfaces, causing an immediate crash on launch.

Notably, while investigating, the agent itself hit the exact same instability: when it reached a “check d8 min-api 26” verification step, it failed and looped through “Agent execution terminated due to error” seven times in a row — the same backend instability pattern described elsewhere in this thread.

This suggests the Build backend produced a genuinely broken build (not just a false quota-exceeded UI message), which then compounds with the quota bug preventing any retry/fix.

19

Update: I tried the built-in “Fix Dependency Crash” quick-action suggestion that AI Studio itself surfaced for this project. It doesn’t even start — stuck indefinitely at “Running for 0s” / “Channeling”, never progresses past that point.

So at this point every path I’ve tried in Build mode is blocked: normal prompts, new chats, different models, and now even the platform’s own suggested auto-fix action. This is currently blocking a production app used daily by our warehouse team, so any update on timeline or workaround would be really appreciated.

20

@Mahesh_Sutar

This is currently blocking a production Android app used daily by our warehouse team. A build generated on Sept 15 produced corrupted bytecode (invalid interface methods in classes8.dex, confirmed via dexdump), and the app now crashes immediately on launch for all users. I can no longer use Build mode at all to fix it — every attempt gets stuck indefinitely at “Running for 0s” (including the platform’s own “Fix Dependency Crash” suggestion), separate from the quota error others are seeing.

Details:
Project ID: inoutfarma-rececoes-497613
Tier: Free
App: “Registo de Receção”

Happy to share any further logs, screenshots or details you need. Would really appreciate a look given the production impact.