I got 112h in the pro plan.
I mean, what in the god’s name is going on with AG?? Why is it not being fixed??
Drops new amazing model. Everybody uses at same time. Migration doesn’t go smooth. People hit turbulence for a few days. Life goes on.
When I was in systems, we never deployed new releases immediately–because there were always bugs. Let someone else wrestle with ‘em. Wait a few days for the bugs to get patched.
I have the same issue. I leave Antigravity there for the whole night but I didn’t do anything, the quota reduced from 80% to 20%.
I also need to know how to check the quota for the models, I too am on Pro plan have not faced this problem yet, as I only use gravity if I hit a road block or fix small errors
Thank you for sharing your feedback.
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! See Google Antigravity for the latest.
okay, but WHAT ARE THE ACTUAL LIMITS?!?!
similar with my case.
I’m in the same boat.
The weekly limit reset on Saturday, yet I’ve barely used it this week, and Gemini 3.1 PRO has already dropped to 20% remaining.
Now I have to wait until Saturday again.
The allowance is clearly decreasing.
Don’t use 3rd party plugin or extension that could burn your token quota. I have no problem with the quota as I don’t have cockpit installed. Cockpit or quota monitor can burn your plan quota without you know it 
And I thought I was going crazy! “Thank god” it’s not just me..
Just this morning I used Gemini 3.1 Pro for roughly an hour? Before receiving an 8-hour ban! How is it even possible?! What is this ratio?! The quota limits don’t make any sense..
I was banned on a Claude model for 7-days! An entire week! The best part? I barely used the model, perhaps just for a few hours (if you could call it “usage”). I was constantly being hit with network errors (each prompt and sometimes even multiple times within a single prompt), despite using the tool off peak hours.. Yet, the quota system thought I overstated my welcome.
C’mon guys, Antigravity was such a brilliant tool! But nowadays I am just paying for a service/tool that I can’t even reliably use! Honestly, I wouldn’t even bother if I wasn’t a paying customer (Pro), but this is absolutely unacceptable. The tool is being advertised as a “coding companion/replacement” yet you barely look at it, and you get a quota limitation.
Each month there is an issue.. Either the model doesn’t work (which burns tokens) or you get hit with straight up quota limits discouraging you from using the service/tool, while paying for it (it’s becoming a monthly Google donation rather than an exchange of funds for service).
I can’t even imagine folks using Antigravity for an actual project/s.. I’m just vibe coding a “simple” app for a few friends and me.. Honestly, at this moment I am thinking about getting refunds (at minimum) as I couldn’t really use the tool..
Yeah, I’m pretty devastated (to say the least) by the current experience.. Would’ve shared screenshots if I wasn’t a new member (but you get the picture)..
I’m afraid that page really isn’t very helpful. From the page itself:
This establishes that there are quotas refreshed every five hours, and quotas refreshed weekly. It also establishes that the most expensive tier receives more of each than the others.
Does that mean that AI Ultra users can expect one more token worth of input/output than Pro users get, or one trillion?
The frustration at the moment over the weekly quotas is that there is no way for us to see how much usage we have left and budget accordingly. Couple that with obnoxious bugs where the model tries to run shell commands and they’re constantly hanging, causing numerous repeated requests, etc and it all gets very frustrating.
Personally I’d be happy to pay for AI Ultra if I had some assurance that it would be sufficient for my daily use, but there’s absolutely no way for me to evaluate whether that’s the case.
This is unacceptable! i’m PRO user and i have to wait days!!!
Where are you GOOGLE? Give us an answer and fix it immediatly!
Wth is going on? I’m a pro user and didn’t use the model for like 12h, then now I just open AG and suddenly my Gemini 3.1 quota is 20%?
I am facing the same issue even writing a small amount of code it shows quota reached
Just to add: tried Codex with GPT-5.3 and it actually burns limits slower than Gemini 3.1 Pro — and that’s on a FREE plan. Gemini is on a Pro subscription, by the way. Limits are weekly, but Gemini’s feel almost weekly too. The workflow difference is pretty obvious.
I am on the ULTRA plan and the same thing – my quota dropped from 100% down to 20% then to 0% within a few hours while I was not doing much of anything. I am not a “heavy” user yet I am being told I have to wait 6 days to use Claude Opus again for my tasks?! How do Google expect to have professional developers use their tools. Can you imagine I tell my boss, sorry no code for the next 6 days because Google unfairly decided I am using too much of their tool. Honestly, are you just trying to make developers go away and not come back?! I need to know if this is a bug that will be fixed or do I need to cancel this plan.
I have AI Ultra and get a warning of “Model quota reached Your plan’s baseline quota will refresh on 3/2/2026, 1:13:39 PM”
All I have been doing is editing an a prototype html/css/js and apparently used everything (including a gap of not touching it for 1 hour or so)? I somehow even used all the GPT quota without ever using it.
Gemini Ultra user here too and I also got locked out until 3/2/2026. 7 days? This has to be an error. Also the speed on the other models is so throttled that it’s unusable. Obviously if this isn’t an error then for sure I’m not going to continue to pay $200 a month to get hit with a 7 day ban when I have deadlines. Really lame. My quota looks the same as yours. I was working on a project this week but far less than weeks prior. I hope this isn’t Google’s intention.