I have used significantly more of the AI model in the past than I did during the three days when Antigravity reported that I had already exhausted my quota. Because of that, I don’t understand how or when this usage is being counted.
In previous weeks, I consumed far more tokens without hitting any limits, so what exactly caused the weekly quota to be reached this time?
Is Google investigating this issue? I have seen other users reporting similar experiences.
Since I am paying for the Pro plan, what guarantees are there that Google is not dynamically adjusting limits behind the scenes? For example, if Ultra plan users require more resources, how can customers be sure that resources allocated to lower-tier plans are not being reduced to accommodate them?
Greater transparency regarding usage calculations and any changes to quotas would help build trust and avoid confusion.
I used to have pro and it worked perfectly fine, my free plan right now is horrible and I can’t work with it. I planned to go back to pro today, but seeing that people are having problems with the quota on the pro plan even scares me away from it.. Is it really that bad now?
this is soooooo frustating… while I understand that it’s basically IMPOSSIBLE to display an live view of token usage for millions of people, I would like to see something like this (can be calculated later):
" worked for 12s , used 2K tokens. [total conversation token: 5M]" on this space here:
Same problem! I suddenly got banned for 3 days! For no apparent reason. The developers have 100% screwed up again with their quota/limit and request calculations. The most annoying thing is that there’s no way to properly track your resource usage; you’re simply forced to take these guys’ word for it, even if they’ve hit their weekly limit, and nothing works! Okay, even if I’ve reached my weekly limit, the week ends on SUNDAY, so why the hell do I have to wait another Monday and Tuesday? Where’s the fucking logic? None of the developers are responding on this forum. It feels like Google has simply given up on people and is committing fraud. When will someone in the US start suing them collectively for this outrage?
Yes, everything is possible and real. Cloud Code has had a transparent system for a thousand years where you can see how many tokens your requests consume and how the percentage scale moves as you use them. Google simply doesn’t care who bought this product.
I have an ULTRA plan and the same thing happened to me. This is my last month paying for this plan. I am going to Claude code! The last 2 months is embarrassing what Google did with all Gemini Pro and Gemini Ultra subscribers.
The exact same thing just happened to me. I have 7 MD files on making sure tokens aren’t vaporized and yet TWO unity UI’s just blasted my entire weeks quota, I’ve filed a complaint and will cancel and move to codex, along with the impoverished image and video generation I feel the google eco system is completely worthless now.
Couldn’t agree more. Even if Google doesn’t want to show token consumption for each prompt, a weekly usage bar—similar to the hourly limit tracker—would go a long way. Users would have better visibility into their remaining usage and could plan accordingly, rather than suddenly hitting a limit with no prior indication.
The whole experience over the last few months has been getting more and more frustrating. I have been very careful to try and guide things to use less tokens and been removing everything I can so it does not get used in the context. I am even trying to use Antigravity less overall so I do not hit quotas.
Even then, yesterday I must have hit a weekly limit and now cannot use it again until Wednesday. While I appreciate there needs to be limits, it it hugely frustrating to not see how you are burning through the weekly ones, as it makes it impossible to plan. Trying to slow down usage for me would be better than having x number of days not being able to use at all. However, I cannot adapt without having any insight into usage impact.
+1 Here, I got hit with this and now I’m locked-out (2 days and 20 hours) after I left it at 60% on Friday night. Must be something on Google’s end, maintenance stuff probably? (I’m fine if true, BUT NOTICE IT FIRST!) Hope they can fix it before Monday 8 AM (-4 UTC).
Its obviously bugged, I’m guessing a disgruntled employee on Friday evening caused a massive disruption by dropping the quota levels for everyone worldwide, its ALL over X and Facebook. This will seriously harm google (intentional?).
I experienced the same thing on version 2 >, I finally found a solution by deleting all the default packaged Plugins. The ones that consume a lot of tokens are the skills in Modern Web Guidance, as much as possible, make your own skills
Same issue her as reported by many of my fellow AI pro subscribers.
Suddenly, almost out of the blue, out of quota for several days. This looks suspiciously like the extremely restrictive quotas that were introduced with Antigravity 2.0 but then quickly rolled back. I’ll give you some leeway this time as well, but if this is the new reality - I’m closing the door.
Are there even people on this forum who have the authority to officially respond to people on behalf of the company? Why are users on every forum thread just talking to themselves? Hey, moderators, what’s going on here? Where are our official responses?
It’s still usable, if you’re using the IDE and not the Agentic application. But at this point, you’re better off just going “back” to VSCode with Claude Code and Codex extensions. I still use the Antigravity IDE myself for now.
I see you’re running a lot of high and medium reasoning, I’m curious what sorts of tasks are these? How do you determine the reasoning required for any given task?
Based on what Ive learned recently, the main cause of quota capping is:
A) Wrong model selection (in this case model/reasoning ) you can literally succeed with 2.5 flash for vision, summary, compression, triage, retrieval etc.
…But the greatest effect on token burn,
B) Context bloat. I noticed when asking Gemini to write system prompts, it was repeating the same statements over and over and over. You can’t even see the entire text when you paste into the prompt field. So I go over to a note pad, and I was completely blown away by how much waste is built in. Any way. You need to be breaking your project into sections, merge often and run fresh sessions at least once a day. Twice is better.
Anyway. I see a lot of people experiencing issues. So something must’ve changed. Still. Hope this helps. The subsidies will end eventually. We have to learn to build concervitively.
Learned that you want flash on low so skip the reasoning for pure coding tasks unless you need to do a tech plan or discuss going forward. Had better results with zed and open code big pickle and it fixed all the slop Gemini wrote on high reasoning. Stuff I never asked for and constantly wanting to read my .env
