While everyone is expecting Gemini AI 3.5 to be released tomorrow, May 19, at Google I/O 2026, some changes took place in advance.

Starting on May 17, 2026, there are some Changes to Gemini model access and limits:

The changes listed below only apply if you are 18 or over. If you are under the age of 18, there are currently no changes to your usage limits.

Usage limits for Gemini Apps

Gemini will move to compute-based usage limits that will refresh every 5 hours until you reach your weekly limit. Calculation of your usage will factor in the complexity of your prompt, the features you use, and the length of your chat. Paid users have higher limits than users without a Google AI subscription.

Premium models and features require more usage and may cause you to reach your limit faster. This would include things like:

  • Media generation
    • Images, videos and music
  • Deep Research
  • Pro Model
  • Extended thinking and Deep Think

Usage limits

PlanLimit
Without a planStandard limits
AI Plus2x higher than standard limits
AI Pro4x higher than standard limits
AI Ultra20x higher than AI Pro

This is crazy!

What needs to be understood:

  • Some people wrongly understood that underage users have no limits. What happens is that their limits didn’t change, not that they’re infinite! It’s likely that their limits are now higher than for adults, but I expect more severe censorship and a complete ban on certain topics for people under 18. The answering style might be different, too, for accounts of minors.
  • The refresh at 5 hours is now explicitly stated, but I’m not sure that there’s anything new in that.
  • The limits in tokens are still not stated, but the usage calculation is better defined: complexity, features, and length, where the length affects the context.
  • It’s a shame that AI Plus only offers twice as much as the free tier, and AI Pro only twice as much as AI Plus. The problem is that AI Pro costs about 2.75x AI Plus (say, €21.99/mo vs. €7.99/mo), but only offers 2x in usage!
  • To be honest, this is not entirely accurate. The official list of features adds to AI Pro “higher limits in AI Studio, Google Antigravity, and Jules” and “higher limits in Gemini CLI and the Gemini Code Assist IDE extensions, plus $10 in monthly Google Cloud credits from Google Developer Program.” If I’m not wrong, those credits are to be used with an API key, not in everyday OAuth usage. And how much higher are those limits for Gemini CLI and the Gemini Code Assist? It’s anyone’s guess.

That said, it’s probably not the end of the world yet. However, hysteria is already happening on Reddit.

Take this one: Gemini Pro New Limit is Messed Up:

Could believe it this morning, but at 1pm, when my limit was reinitialized, I just sent 3 prompts and bam, can’t use Gemini Pro for the next 5 hours. Came here and learnt that Gemini usage had been restricted.

That’s ridiculous. I’m in the middle of writing a book for high school students. I’m at the end of a chapter, so of course the context is huge with all the exercises it contains. And I’m stuck on 5 prompts every 5 hours!!!!

I’m trying not to panic and find a solution. Here’s all the alternatives I’ve come up with :

  1. Create a new Google Account declaring that I’m under 18 because apparently the new usage limit only apply to 18 years and older.
  2. Go and try DeepSeek. Unsubscribe to Gemini Pro because being blocked after 5 prompts is not manageable.
  3. Try to change the way I use Gemini (instead of writing a whole chapter in one discussion, separate each part in several discussions, remove the personalization parameters).
  4. Try Claude: but I’ve heard that Claude is great for code, not so much for the math work and writing text. Plus, seems like Claude also has set a usage limit.

Thanks to anyone who can help me !

Firstly, this guy is contradicting himself: is the limit 3 Gemini Pro prompts in a 5-hr window, or 5 prompts? In my experience as a free user, it’s always been 3 Gemini Prompts until you’re kicked out!

Oh, wait, he says he’s a Gemini Pro subscriber! Not Plus, but Pro. If the “4x higher than standard limits” means “5 Gemini Pro” prompts, then the idiot is using an entire book chapter as context, and that chapter is huge!

I’m not sure why is he that sure that DeepSeek would give him truly unlimited usage. Another guy is also a DeepSeek fan: “I use DeepSeek almost exclusively now. The web client has no limits. The API is cheap af.” Oh, well, as long as he doesn’t hit any stupid censorship… But unspecified (“unlimited”) web client usage is also offered by Qwen.

As for an underage Google account, I’m sure it would be hit by severe censorship.

In other news, some people are shown different options in the Android app. Different from the standard ones, which is what I see in my app. (Source 1, Source 2.)

They both seem to be using paying accounts, but some users of free accounts have received the first change, too. The normal choices for most people (even paid accounts) are still Fast, Thinking, and Pro.

The adjustment of the thinking level would be nice. It would make even more sense in the context of the second change, which might prepare the ground for this future set of options:

  • Fast → 3.1 Flash Lite
  • Thinking → 3 Flash with adjustable thinking
  • Pro → 3.5 Pro

Only time will tell. But people, even if they pay (OK, it’s their money), rely too much on AI. They really deserve that decrease in limits or increase in costs.

Oh, someone has both new changes present, and it seems that the thinking level can be adjusted globally, although it makes no sense for Fast or 3.1 Flash Lite. Maybe it only applies to the other two models. A free user from Bangladesh: