I finally lost patience and left another comment on the forum. They keep making the exact same mistakes rather than letting it be. ;;
Hey yeah, I seem to have encountered this issue as well recently. What do I do about it? Or is this permanent?
No, It’s likely not permanent, but it is a repeated mistake on their end. We actually dealt with this exact same issue for most of the month (as you can see on this thread). They finally patched it around September 23rd, and the limits went completely back to normal; but the new “daily usage limits” update they just rolled out today seems to have completely broken the subscription verification all over again.
As for what to do, just reply directly to the moderator’s recent announcement that was linked here. State exactly which model you were using and roughly how many prompts you sent before it locked you out. Since they just pushed this update and are actively asking for feedback, getting everyone to report the regression right now is the fastest way to force them to patch it again. Hang in there; we mostly just have to wait for them to clean up whatever they broke today.
Thanks, I’ll be sure to do that some later time.
Its more like they can’t leave it alone because running heavy users’ context windows is costing them a fortune in GPU compute. they basically decided they’re done subsidizing that cost on a standard Pro plan, so they built a wall
Everything else about the higher limits seems to be corpo-speak
If they realized the standard Pro subscription can no longer support heavy context windows, they should just introduce a transparent usage dashboard. (exg, Claude with the rolling caps and usage boards,) Give us a visible token budget, clear RPD meters, etc so professionals can actually manage their workloads. Yet they still insist on relying on these completely opaque 24 lockouts that just drop like a guillotine mid-session. If their ultimate goal is to push high-volume users onto the Pay-As-You-Go API, they need to be upfront about the transition, rather than just passively aggressively insisting upon it (along with their troubleshooting ideas like changing models every second) or them quietly degrading the subscription experience until people give up and migrate.
GCP (Google Cloud) has all the usage meters and token budgets you could ever want, because that’s where they actually charge you for the compute. I doubt they’ll do that for studios tho since the ambiguity is exactly how they’re saving money. they know exactly what they’re doing by keeping us in the dark. If you want guaranteed compute, they expect you to pay per token 
Oh, here we go again…
Yes, I can confirm that the quota has dropped extremely hard again. Tonight, I was able to get no more than around 8–10 generations, even though I genuinely thought the issue had finally been fixed. My chat was at about 515k tokens.
At this point, I honestly don’t know whether Google is doing this intentionally or not. But as a paying customer who uses the service in good faith, all I really want is some kind of official statement or response from them. I don’t even care anymore what exactly it says. It could be confirmation that these extreme limits are intentional, or that this is a bug, or even that serving very large contexts is simply too expensive or not worthwhile for them, as someone mentioned above. Just something.
Personally, I’m simply exhausted. These issues are genuinely disruptive and have a direct impact on a person’s emotional well-being. One moment I finally relax after seeing reports from users saying that things have stabilized for them, and then, almost literally the same day I finally get the chance to test it myself, everything breaks again. Apparently, this is some kind of curse 
Because everything is left in this constant state of uncertainty, I don’t know what to do anymore, and I’m literally afraid to send an extra message in Google AI Studio because I don’t know how much quota I have left.
If we just had an official response, or even a general statement, I could at least understand whether I actually have a future on this platform or whether I need to start looking for an alternative again.
At this point, I don’t even want stability as much as I want clarity about what is actually happening.
Please submit a report about your issue here: Daily usage limits for models in AI Studio
Since Google collects feedback there, if this is a bug, there’s a small chance they’ll fix it faster this time.
If this is a result of the policy change regarding usage limits for models in Playground, then perhaps they’ll make the limit calculations for Pro and Ultra subscribers more reasonable (because if 5 requests to Gemini 3.1 Pro in a long chat is a “much higher limit,” it’s scary to imagine what the situation is like for users without a subscription).
By the way, here’s a small observation from the time when this issue with limits was occurring prior to September 23, 2026. This observation remains relevant even now, after September 29, 2026.
When I hit the model’s limit in a long chat with a large number of tokens (500,000 or more), I tried the following to check whether this was a daily token limit:
-
I created a new empty chat
-
Uploaded a large PDF (about 500,000 tokens) or video (about 600,000 tokens)
-
Asked the model to summarize the PDF or video
-
The model responded normally
From this, I draw two possible conclusions:
-
If tokens in videos and PDFs are processed differently than tokens we enter into the input field, then the limit applies only to tokens entered by the user manually, which explains why the limit is exceeded specifically in long chats with a lot of context
-
If tokens in videos and PDFs are processed the same way as the tokens we enter, then perhaps the processing of long chats specifically is failing somewhere
But this doesn’t explain one fact: based on Google’s announcement from September 29, 2026, we can conclude that if the model usage limit is reached, it will be impossible to use the model at all for the remainder of the day. However, even when the limit is reached in long chats, the model can still be used in new chats, which gives hope that this is indeed a bug and not a feature.
Unlurking 2 give my thoughts out on this, but is it possible the usage caps from early last month weren’t actually a bug? like what if it was just a canary release for these new limits? then they temporarily rolled it back on the twenty third to tweak performance metrics after everyone complained, and now the announcement was the official launch.
I might be reaching here, but yea just my thoughts.
Hello everyone again, for a while this rate limit issue had been fixed for me and had returned to normal, but for a few days again with about 10 to 20 few requests I get rate limited on the Gemini 3.1 pro model and this problem has returned again completely on the Pro subscription and it is really disappointing!
Well, if they were testing the new limit system until September 23, they rolled it out on September 29 without any changes.
By the way, I think there were some people here with Ultra subscriptions, if I’m not mistaken? I wonder if they could tell us how their limits have been working out since September 29?
Yeah.. the actual numbers probably didn’t change at all from the shadow-drop. I think the only real difference with the 29th rollout is the PR spin. They’re just saying it’s official now rather than some bug
Yeah. I have Ultra. It feels like I have more than I did before but it’s hard to tell because I upgraded right around the time the limits started getting messed up.
I think everything is tied to the high load on computing resources. The previous occurrence lined up with the 3.8 Flash rollout, and currently, Gemini 4 Argon has launched in limited access. It’s likely that rate limits for ultra-long chats are being artificially restricted during these times to save resources.
Their servers definitely take a massive hit every time a new model drops but I don’t think Google views this capacity issue as temporary anymore; given the pinned announcement,
it really feels like they’re trying to push flat-rate UI users over to pay-as-you-go APIs
Providing a brief status update for the thread’s data log. I realize that documenting this feels a bit redundant following the recent posts, but I still want to maintain a record of the throttled capacity we are experiencing. Today, my account hit the rate limit wall after exactly 6 responses. So, safe to say it has fully reverted to the single-digit caps I experienced prior to the temporary lift on September 23rd. (To be clear, my prompts haven’t been altered at all, no new context has been added, and these were regenerations.) Following the update on 29 September, the limit temporarily seemed to sit at 10+, but as of today, single digits. Back to inconsistency, I guess.
If anyone else is still tracking their daily limits, it would be somewhat helpful to know if you are also seeing hard lockouts under 10 requests again today.
Unfortunately, I wasn’t able to check my limits on September 30, October 1, and October 2, since I wasn’t using AI Studio heavily on those days. That said, I wouldn’t have noticed the issue on September 29 if I hadn’t hit the limit after 5 messages, some of which resulted in an “Internal Error.”
And I think it wouldn’t hurt to document the limit issues in the pinned thread with the announcement—especially since Mustan Lokhand, who posted the announcement, specifically asked us to document any limit issues if we feel they’re too low. Although I don’t really believe this will help if the change was intentional.
Hi, everyone. How are things going with your limits today?