These new limits are complete .... and a massive downgrade for Pro users

Google Antigravity
1

As a paying Google AI Pro subscriber, I’m extremely disappointed with the ridiculous new usage limits you’ve just rolled out.

The limits are so tight they’re making the platform almost unusable for any serious project. What was the point of marketing this as a powerful developer platform if you’re going to kneecap it like this? Other AI services (even some free tiers) are more generous than this.

Google, if you’re going to charge Pro prices, deliver Pro performance and capacity. These arbitrary limits are frustrating, anti-developer, and straight up disrespectful to your paying customers.

Fix this nonsense or watch people migrate to competitors who actually value heavy users.

2

Agree agree agree agree agree…

3

I’m getting ready to go back to Kiro.

4

The limits for Pro users are good now (and they did reset for everyone), they got updated. I wouldn’t expect anything more from them, but this is workable amount. :slight_smile:

If you code intensely you will only hit your 5-hour limit after like an hour (that is if you’re using Gemini 3.5 Flash (Medium) model) I didn’t test with the High one, but that probably costs more tokens.

5

I’ve been a subscriber since around March 2025. These new changes are completely out of touch with the customer base I feel. Saw it glad and won’t be subscribing anymore. My normal chat resulted in 20% quota gone in 1 basic message, 1% weekly. It’s basically 8.25x less quota than before along with a 5hr quota block. Definitely a no brainier to unsubscribe

7

Limits for Gemini Flash 3.1 were more or less reasonable for the Pro subscription for occasional users. Now that we only have Gemini Flash 3.5, those limits were drastically downgraded, and as a result, you can quickly end up not being able to use for a week that you are paying for. Luckily these are the last days of my trial period, so for sure I will cancel my Pro subscription. By the way, anyone knows a better alternative for Antigravity with Ultra subscription? - because since this “upgrade” Ultra is the only way to go.

8

Yeah, it’s crazy. Apparently, with 2*30-minute sessions over 5 hours, I’ve used up my entire weekly limit. It now says next refresh in 6 days and 23 hours. If that is not a bug I’ll go back to Claude Code next month.

9

Barely used anything and now i am locked out for a week, as a paid pro user, absolutely pathetic google.

10

Pro account holder ($20/month), today I was coding a react web app using antigravity and 3.5 flash medium. Using the model in exactly the same way I was previously using github copilot pro ($10/month, utilizing a higher effort model) with proper chat sanitization / restarts / efficient prompts. I blew through my 5 hour limit in 3 hours. On the 5 hour refresh I blew through 3/4 of my weekly usage in another 3 hours. At this pace for $20 a month I am only getting ~8 to 9 hours or one work day of agentic coding a week (I’m not sure yet what this will look like over the full month).

That is unacceptably low. When I was using github copilot pro with a higher reasoning model on a comparable effort project I was easily getting 4 to 5 full work days of coding a week before hitting 3/4 of my weekly limit. I never went over my monthly limit.

Hopefully they see the communities take on these reduced limits and take action otherwise this product is no longer for actual developers.

11

I agree with you I am searching about another company this is last month with gemini

12

True. Those new limits are annoying. Before this “update” (downgrade) i was able to use Pro more before. Now its hits a limit after few messages.

If you think about Google AI Pro plan, dont buy it. Its not worth it now.

14

$20 for nothing now. Real huge letdown! I don’t want to go back to the hallucinating Microsoft CoPilot but at least the limits seemed less predatory.

Its all just so obvious they want people to shell out for the $100 plan. I can’t afford that and I won’t pay that for Gemini anyway since I only use it for personal research into modding, Animanga and personal research projects. That’s not something I can justify spending so much money on.

15

Agree, Locked out for ~4 Days, it seems as if google doesn’t want us to use antigravity any more, where are my 5 hour refresh cycles? earlier I could use it for ~2 hours of active development, ~2 hours for planning, debugging, code review etc., it made more sense developing my product, now since I’m locked out for 4 staight days, I’ll have to pivot to some other solution asap if this 4 day reset is not quickly resolved, they declared cycle reset today, but I didnt receive that cycle reset, why?

16

Most expansive subscription and got a week halt on claude :confused:

17

I agree. Bring it back!

The workhorse is needed. You do not always need reasoning. The 3.5 usage limit is gone faster than i type this sentence. Pro 3.1 and flash is a good combo.

18

I switched to an annual Pro subscription just last month, yet now I need to pay for a separate Codex subscription simply to keep my daily workflow running smoothly.

The previously communicated five-hour reset window was changed without prior notice. This change makes me feel misled about the value and expectations associated with this purchase.

19

These new limits feel like more than just a product tweak. To me, they feel like a signal.

Google is clearly not just shipping AI for today. Their own docs now say Gemini Apps use compute-based limits that depend on prompt complexity, model choice, and chat length, and they explicitly announced changes to those limits from May 26. Google also separates higher usage tiers for AI Pro and Ultra, while the Gemini API itself has its own rate limits. So yes, the limits are real, and yes, they are part of the system now.

My honest take is this: in the current AI era, no company can afford to think only in terms of “we have the best model right now.” That advantage disappears fast. Someone else will ship something cheaper, faster, or more open tomorrow. So the real game is not just model quality. It is user trust, ecosystem lock-in, developer loyalty, and long-term attention.

That is why these limits matter so much. A user does not just remember the model. They remember the feeling of being blocked. They remember hitting a wall in the middle of work. They remember when a product starts feeling less generous, less reliable, and less aligned with how they actually build.

And this is where the bigger fear starts. If users slowly move away from Google tools, then Google is not just losing a few prompts. It is losing the daily habit, the workflow gravity, and the future feedback loop that helps a platform stay dominant. In AI, attention is not a side metric. It is the moat.

That is also why I do not think this is just about pricing or usage caps. It is about whether a company can keep its reputation intact while the market moves under its feet. Google still has one of the strongest ecosystems in the world. That part is undeniable. But in AI, being huge is not enough. You also have to feel fast, fair, and actually useful in the hands of real users.

***So my view is simple: limits are not the problem by themselves. The problem is when limits start making people feel that the product is pulling away from them instead of growing with them. Like When I use some my custom frameworks which are so so good for me no matter what "'m working upon so the ecosystem on which I’m currently opencode I wont change that quickly bec I know that even my limits are over but that ecosystem is so good so I can Pay I wont regret to pay there.

But currently many like mass ANTIGRAVITY AND GEMINI-CLI USERS STARTS THINGS BUT END UP DUE TO THE PLATFORMS WEAK STRUCTURES.***

That is how trust gets damaged.
That is how users start exploring alternatives.
And that is how even the biggest names in tech can slowly lose the momentum they thought was permanent.

Maybe Google will adapt. Maybe the ecosystem will settle. Maybe the next few months decide a lot more than we expect. But right now, this move does not feel like confidence.

It feels like pressure.

20

image
image662×293 6.16 KB

Their new 3.5 model is a massive token eater that I can barely finish the average daily tasks on the project, even with their 3x quota increment (Of course, all these tasks can be done within 10 or 20% usage of 4 or 5 hours refresh in the previous Flash 3 model).

First, I was just trying to understand the quota limits, but now I realised it also has the weekly limits?? Then why should we keep paying for the product while other competitors offer better models with better quota limits in the same price zone? And they don’t mess up a lot with their tools they can patch it right away instead of ignoring their customers like these guys, who took a long time to even clean the spam bots because they don’t even check their site.

21

I need to vent, but more importantly, I want to see if other developers are hitting the same brick wall. I’ve been a loyal Google AI Pro subscriber using the Antigravity IDE, but the recent rollout of Gemini 3.5 Flash has completely broken my workflow.

Google pitched 3.5 Flash as a faster, more efficient model optimized for agentic workflows.

22

I see lots of people are not happy with the limits. I understand that until 3 days ago, before the 3x limit was given, I am being able to work for hours without even reaching the 5h limits. Also I worked 12 hours each day from friday to sunday. Tonight they reset the limits for free again. So if you are not feeling the improvment something is really wrong. Some people complains that if you have large files in your project it can affect the consumption very heavily. I was in the same boat as everyone until IO 2026 when things got much better.