I am a Google AI Pro subscriber using Google Antigravity, and I believe there is a bug or undocumented change in how quotas are being applied to my account.
According to the official Antigravity plan documentation, Google AI Pro subscribers should receive “high, generous quota, refreshed every five hours,” with higher weekly rate limits than free users. In other words, the maximum expected wait time after hitting quota for Antigravity usage should be about 5 hours, not multiple days.
However, my current Antigravity UI is showing much longer cooldowns for several models, even though I am on the Pro plan:
Gemini 3.1 Pro (High) – “Refreshes in 9 hours, 43 minutes”
Gemini 3.1 Pro (Low) – “Refreshes in 9 hours, 43 minutes”
Gemini 3 Flash – “Refreshes in 4 hours, 56 minutes”
Claude Sonnet 4.6 (Thinking) – “Refreshes in 18 hours, 10 minutes”
Claude Opus 4.6 (Thinking) – “Refreshes in 18 hours, 10 minutes”
GPT‑OSS 120B (Medium) – “Refreshes in 18 hours, 10 minutes”
This behavior feels very similar to the multi‑day lockouts other Pro subscribers have reported (6–10 day lockouts, 5‑day and 7‑day windows, etc.), where Pro accounts are effectively being treated like free‑tier or weekly‑limited accounts instead of getting the 5‑hour refresh that is advertised.
What I expected
Quota buckets that reset roughly every 5 hours for Antigravity usage on my Pro plan
Higher rate limits than free tier, without long multi‑day lockouts
A maximum cooldown of around 5 hours after hitting quota, as described in the Antigravity plan page and Pro/Ultra announcement blog post
What I’m seeing instead
Per‑model quota panels in Antigravity showing cooldowns of ~9 hours and ~18 hours for multiple models
Behavior consistent with the “secondary weekly cap” / multi‑day lockout pattern described by other Pro users, rather than the documented rolling 5‑hour refresh
No clear in‑product explanation that Pro subscribers can be subject to these much longer cooldown periods
Environment
Plan: Google AI Pro (active)
Product: Google Antigravity
Affected models (so far): Gemini 3.1 Pro (High/Low), Gemini 3 Flash, Claude Sonnet 4.6 (Thinking), Claude Opus 4.6 (Thinking), GPT‑OSS 120B (Medium)
Questions / Requests
Is this behavior (9–18 hour cooldowns on a Pro plan) expected, or is it a bug in how quotas are being applied to my account?
Has the quota policy for Google AI Pro in Antigravity changed from the previously documented “refresh every five hours”? If so, can you please share the updated, official limits?
Could you verify that my account is correctly recognized as a Pro subscriber in the Antigravity quota system (no misclassification or entitlement sync issues)?
If this is a bug, please review and correct my quotas so that they align with the documented Pro‑tier behavior, and consider updating the documentation or in‑product messaging to clarify current limits.
Right now, these longer lockouts significantly reduce the value of the Pro subscription for me, since I chose this plan specifically to avoid multi‑day waits when working on projects inside Antigravity.
Thank you for looking into this. I’m happy to provide timestamps
We appreciate you bringing your concerns regarding the quota to our attention. Please be assured that we have shared this feedback with our internal team.
As we balance giving the best possible quotas and maintaining fairness between users, especially under incredible demand, we’ve established weekly limits for all models. These limits do not apply to Google AI Ultra, which continues to be the best plan for power developers.
Thank you for responding to the inquiry, and for clarifying on behalf of Google that Google AI Pro users will now have their quotas refreshed every 7 days instead of every 5 hours. You are the first official source from Google to announce this, and we appreciate you taking this step on Google’s behalf.
Can you please further qualify exactly what the rate limits are for each model within their 7-day quota windows?
I am also facing the same issue with reset taking days. I am sharing the screenshot. Please suggest how soon it will done. I am a pro member and I am unable to work now. Till yesterday morning would say around 11 AM IST it was all working fine since then I have not used it however today I opened Antigravity and I see the notification of refresh taking days.
I’m having the same issue, quota refresh plan is not working on a 5 hours based refresh as pro plan promised i have a 6 day refresh windows so if use this 40% will have to wait 6 days again ,also the quota for Claude Sonnet, Claude Opus and GPT-OSS was wiped away without even using those models.
Hi everyone, aside from updating my Antigravity, I’ve barely done anything recently. However, I found out today that the cooldown time for my 3.1 pro model has increased to 5 days. I reached out to Google official support again, and the reply I got was as unhelpful as usual.
i have the same issue, i havent use any anti gravity yet today, just launching the program after the 1.20.5 update from 1.20.3. last night. checking around 11 oclock. quit the program after trying to use flash model and got error server qouta exauseted.
I’m experiencing a critical quota sync error. Despite being a Pro subscriber with plenty of usage left yesterday, my IDE is showing a 6-day lockout today. It seems my account has been incorrectly moved to a weekly free-tier cooldown bucket. Is this a known bug in the latest sync, or a silent limit change? If this isn’t resolved, I’ll have to migrate back to Cursor or Codex as I can’t have my workflow halted for a week!
I have experienced exactly the same too. I made sure not to use up my Gemini 3.1 Pro High/Low or Claude allowances. They were all due to reset to full in a few hours. I’ve checked today and they’re each at 5 day lockouts… That means I have only the Flash model to work with, which then burns down quite fast when I’ve been locked out of the other models…
i have 2 accounts of google ai pro plan.1st one showing same problem as you, even worse. t shoing 5 day cool down proble, 2nd acount showing 4-5 hr cool down period. they promised a generous credits in antogravtiy . but looks like , claude code is much cheaper as compared to it
Experiencing a similar 6 day lockout on Gemini 3.1 Pro models. Yesterday I barely used it and was at 100% quota limit and today upon logging in I am at 0% and on a 6 day lockout. Apparently this is a prevalent bug affecting many, is there a resolution timeline on this because waiting 6 days for an unlock is not acceptable.
I’m on Ultra subscription, and I have the same issue. Until now I’ve only encountered 5 hours timeouts and so very rarely, but now after renewing my subscription I suddenly see 24 hours timeout, I believe it’s a bug and can be fixed asap or have my refund processed.
I am writing to voice my extreme frustration regarding the opaque and inconsistent “Baseline Quota” mechanics currently implemented in Antigravity.
As a software developer and a startup founder, I transitioned to the AI Pro plan with the expectation of a reliable, professional workflow—specifically based on the advertised 5-hour rolling refresh window. However, I was just hit with a 6-day lockout, telling me my quota won’t refresh until March 19th.
This is not just a “limitation”; it is a complete workflow killer.
Here are the core issues that need immediate clarification from the Google team:
Transparency of “Work Done”: We are told quotas are based on “Work Done,” but there is no real-time dashboard showing how much “Work Done” an agent task consumes. I hit my weekly baseline in just 45 minutes of active coding. How is a professional developer supposed to plan their week when the tool can become a brick at any moment? “Modified by moderator”
Inconsistency across Models: While my Gemini 3.1 Pro is locked for a week, other models (like Claude and GPT) remain available. This suggests the lockout isn’t about global server capacity, but a specific, aggressive throttling of Gemini users.
We chose Antigravity because of its potential to revolutionize “vibe coding” and agentic workflows. But a tool that locks its users out for a week—without warning or a clear usage meter—is not a tool; it’s a liability.
I sincerely hope the team revisits this “Baseline” logic. Reliability is the most important feature of any IDE. Currently, Antigravity is failing that test.
I noticed this after updating to Antigravity 1.20.5. The lockous are insane from 5-7 days even without exhausting the quota. I stop using model when it reaches 20% then I switch to another.
