I am a Google AI Pro subscriber, but the Antigravity IDE is applying extreme quota limits far beyond the documented 5-hour maximum for my tier. New chats show a 4-day lockout, while existing sessions show a 7-day lockout (until Jan 16).
Steps to Reproduce:
Log in with a Google AI Pro account.
Use Claude Opus 4.5 (Thinking) in an existing project.
Observe the “Model quota limit exceeded” message with an incorrect resume date (1/16/2026).
Open a new chat and observe a different incorrect date (1/13/2026).
I was using it for analysis my code and debugging the JS problems, but I was getting Limited “You have reached quota limit….You can resume using this model at 1/10/2026, 11:32:26 AM“
So it was infact yesterday I encounter this. And by 11:40 AM I resume the uncompleted task, by sending continue command, and after around 15 minutes it says, “You have reached quota limit….You can resume using this model at 1/13/2026, 8:03:80 PM“
5 days refresh instead of 5 hours.
This is particularly for the Claude Sonnet 4.5 and other including GPT-OSS 1208.
Hopefully Google Antigravity Team rectify this soon.
Tengo el mismo problema y no me deja Usar Claude Opus, llega la fecha y hora que me va dejar usar, y luego se desplaza por más tiempo y no me deja usarlo
I just saw the announcement on Antigravity’s official X account.
Apparently, this was intentional; they’ve restricted Pro accounts to the Free Tier limits. Honestly, this is completely ridiculous—what’s the point of subscribing to a ‘Pro’ plan when free accounts have literally the same limits? It’s easier and much cheaper to just create a bunch of free accounts than to pay for nothing.
It’s highly likely they did this to harvest everyone’s project data, since Gemini can see your code beyond just the chat interface. Then there’s the announcement about Gemini 3—the more capable, ‘premium’ model that will only be available on the Ultra plan after being tested by ‘trusted testers’ to ensure ‘safety’ and all that.
In reality, they probably just funneled everyone into Antigravity to collect data and train Gemini even further. Thinking about it now, it makes no sense for them to release a competitor like Claude 4.5 Opus for ‘free.’ So, Google and the DeepMind crew likely did this for mass data collection, scraping not just your chat history but your entire source code.
I’m canceling this plan as soon as I can. Google is definitely not trustworthy, especially after saying one thing and doing another—the worst part is that it’s even in their own documentation regarding the 5-hour renewal.
Their excuse was that only a ‘minority’ was affected, but the recent outcry proves that’s a lie; even those who didn’t ‘abuse’ the system are hitting these limits. Besides, this ‘abuse’ doesn’t even exist—it’s just a form of social engineering. How can you abuse a service by using exactly what was provided? If they say, ‘you can use our service, but if you use it too much, you have to wait,’ that’s not abuse. Everyone is just using it exactly as intended. Real abuse would be hacking the quota system to bypass access, not using it legitimately.
Hello, thank you for flagging the issue.
To ensure our engineering team can investigate and resolve these issues effectively, we highly recommend filing bug reports directly through the Antigravity in-app feedback tool. You can do this by navigating to the top-right corner of the interface, clicking the Feedback icon, and selecting Report Issue.
This is no bug, they have intentionally done this.
Google is always known to do this kind of things. They are restricting the usage of Claude models because they have to pay Antrophic for what we use.
I hit the weekly limit just like you people did in the same time period. I will tell you honestly, Gemini 3 Pro is the worst coding model I have used, it simply tires to over engineer the solutions to the bugs which are very simple.
It checks this and that, simply wasting time. On the other hand Claude Sonnet 4.5 (thinking) the best model which actually acts and codes like senior engineer.
They know their models are bad, very very bad.
I connected with Google’s chat support and I experienced the worst experience ever. I never expected to wait in Queue for chatting.
The person I was connected with was rude, he mentioned that there is no dedicated support team for Antigravity product. When I asked him then why did you advertised 5 hours refresh on the product page while I was subscribing to this?
That was the moment he ended the chat from his side. I have the full transcript of the chat with me. If anybody from Google’s so called support team is reading this. Reach out to me I will share the transcript.
I have been locked out of the usage even after being on AI Pro plan, and I can not handover my code this worst Gemini 3 Pro.
It deletes working code, literally gaslighting it.
I purchased an annual subscription last month. However, I was locked out after 8–10 quots per week.
Gemini 3 Pro model also runs out of quota quickly, making it impossible to handle complex coding tasks. ( go/wait infinity loops, Agent errors without MCP, using the unknown enum/variable/functions, just made itself, unknown language response, etc).
Although I could solve the issue to some extent by cross-reviewing with a second model. If I have to rely on a single tool without complex workarounds, switching to Tool. I’m considering refund.
I’m going to check the Quota limit that will be released tomorrow and decide.
I felt like the basic Quota was also reduced, but in 2 days, after 4 quota, I went into the 5th and I get a message limit until 28, Jan. but Gemini 3 is a lot of Go/wait infinite loop and Agent Error.
The Gemini 3 model is not follow Gemini.md config and even troubles in the task… I requested a full refund.
