I’m on the Ultra tier and I need to talk about what’s actually happening with quotas, because I think it’s being seriously understated in the official communication.
This is not a “change in how credits are structured.” This is a massive reduction in effective limits.
My before/after:
— Before: multiple agents running complex tasks simultaneously for 5+ hours → quota didn’t go below 80%
— After the update: same workflows → 0% in under 90 minutes
That’s not a restructuring. That’s a 5x to 6x cut in what Ultra actually delivers.
The messaging around “AI credits” and “seamless scaling” is designed to make this sound like an upgrade or a neutral transition. It’s not. We’re paying the highest available tier and getting a fraction of what we had before, with no clear explanation, no real-time usage visibility, and no honest acknowledgment that limits were reduced.
If this is intentional, say so. If it’s a bug, fix it. But don’t dress up a downgrade as a feature.
Yea, I am experiencing this as well, as is another developer I work with. Both on AI Ultra. The issue is this forum is flooded with Pro subscribers getting mad about quota, so even if it is a bug I don’t think Google will notice. They just see Pro subscribers complaining that their $25 didn’t buy them unlimited usage.
I don’t even get through half a day of using antigravity without my Claude quota dropping to 0. It doesn’t feel great being met with suggestions of overages when on the ultra plan, after just over an hours use, and the communication on why feels non existent.
To provide you with greater control and a seamless path to scale, we are evolving our Google AI plans by introducing built-in AI credits that can now be utilized directly within Antigravity.
Under this updated structure, our Google AI Pro tier is tailored for practical builders, offering generous limits for Gemini Flash alongside a baseline quota to experience our most advanced premium models. For developers operating at the highest scale who require consistent, high-volume access to our most complex models, Google AI Ultra serves as the ideal comprehensive solution.
Furthermore, we’ve designed these plans with maximum flexibility in mind; if you are on the Pro plan and require additional capacity or deeper premium access for intensive development sprints, you can seamlessly top up your AI credits to customize your limits and continue shipping your best work without interruption.
it’s understandable to introduce overages for AI Pro users @Abhijit_Pramanik , not for AI Ultra. Isn’t that the whole idea and reason for having an ultra plan?
90 minutes of usage on Ultra definitely shouldn’t force overages to keep using it
Ultra plan became unusable, before I could work with no interruption on two or even three projects at once and run out of quota around 30 minutes until refresh, it was almost perfectly balanced. Now it takes 1 project to run out completely in 1 hour and all my monthly AI credits already gone in first week. All my work/hobby projects are completely blocked now.
Not only are resources strictly throttled, but their servers are also persistently returning HTTP 503 Service Unavailable errors, causing frequent interruptions to the ongoing workloads.
This has absolutely nothing to do with our complaint and hilariously proves exactly what I said. Everyone here is talking about AI Ultra, which has been completely destroyed. This says nothing.
This thread is growing and the pattern is clear. Let me summarize what Ultra users are actually reporting:
@swish: two or three simultaneous projects before, now 1 project kills the quota in 1 hour — and monthly AI credits gone in the first week. @kasperrt: 90 minutes on Ultra before hitting overages. On the highest available plan. @Vu_Vi: on top of everything, persistent HTTP 503 errors interrupting active workloads.
My own numbers: before the update, multiple agents running complex tasks for 5+ hours didn’t bring me below 80%. Now I hit 0% in under 90 minutes.
This is not isolated. This is not a misunderstanding of the new billing model. This is a consistent, dramatic reduction in what Ultra actually delivers — and the only official response so far has been a copy-pasted marketing paragraph that doesn’t mention Ultra users at all, let alone address the degradation.
@Abhijit_Pramanik — with respect, the community deserves a direct answer to a direct question: were Ultra limits intentionally reduced? If yes, by how much and why? If no, when will this be fixed?
I am currently subscribed to the Ultra plan. Previously, I would have to run about five concurrent tasks just to come close to exhausting my usage limit. Now, however, even a single intensive task depletes my allowance within three to four hours.
Echoing the experiences of others here, I have noticed a drastic reduction in usable capacity since the most recent update. I sincerely hope this is simply a bug. However, reading through the community discussions, there is speculation that this might be an ‘intentional’ change, seemingly designed to accelerate credit consumption.
If this is indeed an intended adjustment rather than a bug, I strongly request that Google provides a clear, official notice regarding this policy shift. My subscription renewal date is fast approaching, and I need official clarification to decide on my next steps regarding my subscription.
Dude, same issue. I can understand AI quota adjustments because compute is tight right now and there are people “abusing” it. But what makes me angry is that there were no real terms or notices beforehand specifying that quotas would be lowered after a certain time. To me, this is a complete lack of contractual integrity, especially for a company as big as Google.
Honestly, Claude handles this much better; every time they lower or increase quotas, there’s an announcement, even if it’s just a tweet, instead of making those of us who paid real money guess whether it’s a bug or intentional. Especially now that they’ve doubled the usage for users during off-peak hours until the 28th.
My subscription expires at the beginning of April, and if this issue isn’t resolved by then, I won’t consider renewing, and I believe many other developers will do the same.
Ok, this quota refresh is stupid, even Sonnet runs out in less than two hours, what am I supposed to do while waiting for refresh? Given AI credit prices, it looks like service price silently quadrupled with Google expecting us to buy month worth of credits every week.
Just making quota refresh on daily basis would easily make it usable again, giving solid uninterrupted ~10 hours of work each day at this rate, instead of these stupid 3-4 hour breaks with me staring into the void reconsidering my life choices.
I’m going to chime in here as a Pro user that was considering upgrading to Ultra to fix my quota issues. These stark warnings about the status of Ultra subscriptions are disappointing. I have decided to forgo upgrading and spend the next couple of weeks on vacation trying to figure out my new toolkit going forward. It seems clear that it will not be Google Antigravity. It’s a real shame because I think, for once, Google was slightly ahead of the curve on development tools.
I’ve noticed this as well. I’ve been using Ultra for about 6 weeks and never once hit a quota limit. In the last two days I’ve hit the limit each morning after 2-3 hours of work. Previously, I could work on several projects for hours and usage rarely dipped below 60%.
This is frustrating from a product/cost perspective, but also frustrating from a user perspective. If Google is going to drastically reduce quotas, they should at least let users know so we can adjust our workflows accordingly.
Google AI ultra here too. As of this week, the quotas were reduced drastically. Until last week, adopting the same workflow, I rarely dropped below the 60% quota mark at the end of the 5h window. As of now, I am hitting 20% leftover claude quota after ~90-120 minutes.
This is an undocumented change, and it should be clarified if this is inteded behaviour (related to the introduction of ai credits) or if it is a bug. If this is not a bug, I won’t be renewing my subscription next month, as AI Ultra quotas as of now are the same as AI pro in november.
This is a very serious issue for someone paying 279€/month, that is 323$.
Please respond concretely
Same here, before I was able to work all day long, never went under 60%, 40% in heavy dev days. Since this week getting out of quota after 3 hours, with this new message telling me that I can pay more than the 200$ i’m actually already paying to get more quota…. Extremly frustrating and disapointing.
anybody here used claude max 20x? can share if the quota mirrors the current quota for ultra plan? I was considering it vs renewing my ultra sub but ended up renewing my sub for another month so I can finish my projects.
According to my research, even with Claude Max 20x there is a weekly quota and if 3-4 projects were running in parallel that weekly quota can also get hit fast if using opus 4.6 mainly. Online estimates suggest the following limits:
~900 messages/5h (or 200–800 Claude Code-style prompts) 240–480 hours of Sonnet 4.6 per week 24–40 hours of Opus 4.6 per week
It feels like if we use opus 4.6 with 2 concurrent editors, we might finish the quota within 90-120 mins per 5 hr cycle. so assuming 2 hrs per 5 hr refresh cycle. one day 24/5 * 2 = 9.6 hrs, round down to 9 hrs. 1 week = 63 hours.
I read online that even if you sub directly under Max 20x, quota have also reduced since earlier days. Honeymoon period over I guess.
may want to reconsider - someone at reddit found yesterday that in Antigravity Opus thinking budget is set to the lowest possible token number at 1k (out of 120k possible).
Which would actually explain why Opus in Antigravity tends to drift from specs quite quickly
It’s a real shame because I think, for once, Google was slightly ahead of the curve on development tools.