No one from google has given an official statement, no mail to subscribers, no post on official forum. The real question is, is this a bug or just the new claude quotas ?
I sincerely doubt that this is a bug, it’s like a week (or more) that this thing it’s going on. There was a random reply on x from someone from google about new strict limit for ai studio, so i think this is just the quotas being reduced.
We just need some info, “sorry for this, we have working to solve this bug” or “this is the new quota since the traffic is too high” or something like that.
We can’t use a tool that we don’t know if it’s gonna work tomorrow, a tool that we are paying for. We actually don’t know what we get for what we are paying (don’t mention the fact that we subscribed to a service that has changed during our subscription. I’m not a legal but…)
So, please google, just let us know what’s going on. I can’t get how such a big company doesn’t give informations about their paying services
I suspect both the auth issues, and quota reset issues are not really technical issues but an attempt to keep the concurrent users low by faking technical issues. But even with that they don’t have the capacity to serve those who attempt inference.
Probably to entice people to buy the Ultra plan (since Pro is free for first month). I wonder if Free plan users also get no or less “agent crashed” errors when using claude models (paid plans get it 90% of the time currently). Bait n switch.
They put weekly limit without any information. They announced this on X (link to the tweet) and there is no way to monitor them, it’s like Russian roulette, you plan to do stuff, and the day after you have to wait 5 days to be able to use the model…
Thank you for your continued patience as we work through the high demand for Antigravity.
As @JuCho pointed out by sharing the update from the Antigravity team on X, we have established weekly usage limits across our models to ensure equitable access and platform stability during periods of incredible demand.
For a full breakdown of the different tiers and how quotas are managed, please refer to our official documentation: https://antigravity.google/docs/plans.
Ok great, 5H work 5 days in the box it is just great from you and where i can see mu quata for the week ? how i know how much from the quata i have? For paid service is just not ok.
Okay, but honestly, the lack of visibility into weekly usage makes it extremely challenging to manage the work. I usually prefer to plan, but without this information, it’s now impossible for me to do so !
Can we have a visibility on it? thx.
I just hit the limit after a 57-minute coding session where I only used a few prompts. The system is now telling me I have to wait 4 hours (until 19:32) to resume.
I’m a freelancer and I explicitly switched my workflow from JetBrains to Google’s ecosystem to utilize these models. I am paying for the Premium subscription, but honestly, I have zero use for the bundled 2TB storage or other Google apps — I just need the AI to actually work as advertised.
I am a solo developer working on my own projects, not a large corporation; I cannot afford a €200+ monthly Enterprise plan just to get basic coding help. It feels like a bait-and-switch.
Does anyone know a solid alternative where I can pay strictly for a reliable AI coding assistant without being forced into a bundle of Google services that throttle my usage like this?
You could switch to Claude Pro. It’s about $20 per month and works well for regular coding. If you still hit limits, some users just use two Pro accounts, which comes to around $40 per month total and gives you roughly double the usage.
If you code heavily all day, there’s also Claude Max for about $100 per month. That plan gives much higher limits and feels closer to unlimited compared to the cheaper tiers.
Another cheaper route is using GLM models. They’re pay-as-you-go, and many solo developers end up spending around $5–$10 per month depending on usage. You can also integrate GLM into Claude Code pretty easily. The coding performance is actually pretty decent for the price, and in many cases it feels more reliable than the current dumb Gemini models.
So in simple terms:
$20/month → Claude Pro
$40/month → two Claude Pro accounts
$100/month → Claude Max (heavy use)
$5–$10/month → GLM models
Any of these will be more predictable than getting locked out for hours in the middle of a coding session.
Thanks for the breakdown, definitely good to know!
For now, I’ve actually decided to give Cursor a try. It’s also $20/month, but what sold me is that they have reasonable higher-tier plans (like Pro+ for $60) if I ever need to scale up, which is still very affordable for a freelancer compared to Google’s enterprise pricing.
The migration was insane - it took literally 2 minutes. Since Cursor is also built on VS Code (just like Antigravity), all my keybindings and extensions transferred over instantly. Going to test it out for a few days to see if it handles heavy sessions better!
