So Pro plan users are now paying for Gemini Flash + 1000 credits (9 prompts)?!

[seaheachxv]

I have been using AG for 4 months now, and I’ve never encountered multi-day lockouts, but from last week, and particularly the last few days, I’ve been hitting limits and exhausting quotas way faster. For the first time, I got a 3-day lockout (weird, but still usable with Flash), but after the 3 days, it jumped to 7 days???!!

Antigravity’s LATEST post: “Google AI Pro is the home for the practical builder, hobbyists, students, and developers who live in the IDE and don’t necessarily rely on an agent. This plan features generous limits for Gemini Flash, with a baseline quota included to “taste test” our most advanced premium models.“

Does this mean that access to 3.1 Pro is effectively gone, or is it now an occasional, higher-tier model access??!? It appears we are now paying for a “High, generous quota, refreshed every five hours until weekly limit reached” and “Higher weekly rate limit” for the Flash model instead, with no access to 3.1 Pro, Claude, or GPT. And now we’re just supposed to “taste test“ the higher-tier models?? Link: https://x.com/antigravity/status/2031835833716625883

I tried using the AI credits, and in reality, from the 2 or so days I’ve tested across two Pro accounts, you burn through 80-100 for semi-light tasks and 200+ for heavier tasks, which gives you 5-9 prompts per month unless you choose to pay for credits.

I would’ve much preferred, instead of advertising that the new “Pro” plan is the home for the practical builder, you just straight up announced the new changes as they are, and as we are experiencing them.

Have you tried using Flash for good/real development?? Flash rarely follows instructions well and jumps into tasks too fast, seems to make its own decisions that 3/3.1 rarely made, only to apologise for the mistakes it’s always making; it’s only good for small tasks or larger tasks broken down into smaller ones.

Just tell us you’re gutting the Pro plan. I would’ve much rathered that than having to be met with surprises that literally have not happened in the 4 months I’ve used, and loved your product.

I don’t think I can use AG for real development anymore without jumping to the Ultra plan (which is super expensive). And the credits in the long-term do not seem worth it at the moment.

.

[Leo5]

9 prompts? lmao. I’d say 3 max
yes , you got it right. And the app is crashing at a rate per hour that i stopped counting as it is really unbeleivable they busy shipping paywalls but not bothered fixing bugs on this app

[John_Christian_Lonni]

I did two Claude Opus requests that edited some 5 files in total (not many changes). It gobbled up 500 credits - all in 15 minutes. As it is now the Pro subscription = 30 minutes of Claude Opus. No doubt the other models are cheaper, but even those would hardly last a day I think. AI Ultra 25000 credits at $250 a month would be used up in 6 hours time (4000 credits an hour).

[Vu_Vi]

If you upgrade to the Ultra plan, the limits are almost untouchable for standard programming tasks. I upgraded my account from Pro to Ultra and the difference is night and day. They might be intentionally throttling Pro users.

[Ipinrere_Oluleke-Oke]

True, but I’m telling you, you’re better off paying twenty bucks for a Codex subscription. I know they’re running a 2× promo, but what you get from Codex is way better than anything out of Gemini right now. The more I use 3 Flash the more I start to love it, even though it’s high-key dumber than 3.1 Pro, it’s just dumbfounding, man.

The problem was obvious: they wanted everyone on the same platform, so they set a low barrier hoping people would migrate to Ultra by the time they pulled this stuff. Guess it’s working, but I’d rather cancel my AI Pro plan and move to Codex because it just works. I’ve never given 5.4 a prompt and had it mess up something simple like types.

[Raymer]

Yea I moved to Claude about a month ago. In fact I think tomorrow is exactly a month. And it’s been the best month, much higher rate limits .. not this where I have to sit for days waiting to do anything. And the price is exactly the same. 20 bucks a month…

[YUSUFUCAR]

When my AG subscription ends this month, I’ll be switching to Claude.

[coelhodev]

I think the last release fixed that: Google Antigravity

[Sebastien_Rochon]

I just coded a complete CRM for my company using Claude code pro 17$ plan used only 88% of my current weeks tokens. Had Gemini pro 27$ do 10 actions 4 page pdf a few code reviews and 2 html email templates for 1 week cooldown and stuck with flash only. Up your game Google I will also be cancelling and living with claude code pro agents that dont show up in your agents tabs but do the same work.

[John_Christian_Lonni]

$25 to $250 a month cost is a massive jump for a hobby programmer who wants to use AI assisted development. While I can understand that for a business $250 is nothing, but I feel they have just nerfed it way too much as it is now for normal hobby developers.

[Shivam_Singh2]

Hello Everyone,

Thank you for bringing these concerns to our attention. Please be assured that I have shared your feedback with our internal team for further review.

Thanks!

[Phil_McCool]

Yes, all the noise about a bug and fixes was simply noise. It is not a bug it is a feature. Google has crippled its AI Pro offer with new terms and conditions without announcement. 1000 AI credits per month is of no use to anyone. That monthly limit is why folk are seeing perpetual model lock outs - it is not a bug it is a policy. Bye bye Google AI. Hello Claude.

[Phil_McCool]

Same experience here. Made my decision for me

[Lucaso]

Just move to Codex or Claude, much better and faters.. Even Copilot is better and cheaper..

[Mohamed_salah1]

this is not true now, ULTRA now has the old pro limits

[teef]

Does it really make sense now to have an AI Pro Plan at all?

Claude Pro now seems better for cheaper?!

[WesleyB]

Coming from Codex, I’m genuinely disappointed.

I subscribed to the Google AI Pro yearly plan yesterday, and a single “Planning” prompt using Gemini 3.1 Pro (High) - which took ~10 minutes to complete - already consumed 60% of my 5-hour rolling quota, leaving me with just 40% remaining.

That ratio feels disproportionately high. My intuition is that a comparable prompt in Codex using GPT-5.3-Codex (High or even Extra High) would have used at most 10% of an equivalent time-based quota.

Another concern: the “Planning” task actually modified files in my workspace - something that has never happened with Codex, even on complex workflows. This unexpected behavior, combined with the steep quota cost, makes me cautious about using this model and Antigravity for iterative development.

Beyond the consumption rate, the lack of visibility into weekly quota allocation makes me hesitant to experiment with the more capable models.

I’m aware of Codex’s recent promotional offers, but even accounting for that, the value proposition here feels very unclear. I’m frustrated with the investment so far. I’m hoping that the other AI tools included in the Pro package will make up for it.

Has anyone else noticed similar quota behavior or file-modification surprises? Or am I missing something in how these limits or safeguards are configured?

[YNd]

@WesleyB

Has anyone else noticed similar quota behavior or file-modification surprises? Or am I missing something in how these limits or safeguards are configured?

the whole forum here noticed it

I’m an Ultra plan subscriber, quota drop suddenly and drastically appx 25 of march, made Ultra plan completely unusable for serious work. Since then google support feed me with «we are enhancing Antigravity experience». It turns out the «enhanced experience» is weekly quota and broken dumb models

Just switch back to codex, or use cursor, claude or whatever company, with primary target to developers. Google will gift for free «Pro» subscriptions to the whole world, but they don’t care about paid users, even if it top tier subscribers

[WesleyB]

wow , I don’t get it. Since I posted, I started using only Gemini 3 Flash, and tried Claude Sonnet just once, and now I see the Gemini 3.1 Pro (High/Low) Refreshes in 6 days, 2 hours, in less than the 5 hour time window!?

[cardonator]

Yep, this is pretty typical at this point. AGY is essentially completely unusable without paying per credit, which gets expensive fast. Gemini CLI is better but they have also neutered that to where you get 5 hour refresh once then it goes to 24 hours on the next refresh.

These tools were good but for whatever reason they have decided to make them garbage. I will NEVER subscribe to Ultra. The only reason I don’t cancel Pro is because the majority of the money I’m spending is for Google One features, AI is just an addon. If it was the primary thing I’m paying for,I would have canceled a month ago.

And the only reason they are getting away with this at all anyway is because Claude also ruined their quotas recently so Pro is practically the same.