Earlier this summer, Google rolled out a series of upgrades to its Gemini AI apps, making them more powerful than ever. However, with these upgrades come new limits on how much you can use the service.

Google now measures Gemini AI usage by the computing power required by your requests, rather than the number of requests made. This means that while before you might have been able to generate three videos a day, now you might find yourself limited to just two if the videos are particularly complex.

The new system is more resource-efficient for Google but less predictable for users. The company’s support docs mention that access may be subject to change based on testing and experimentation, meaning that some days may offer different levels of AI prompting than others.

Users can manage their Gemini AI models from the prompt box in the app, with each model having different thinking levels—Standard, Extended, and Deep Think—which affect response quality, speed, and usage limits. The context window size also varies between plans, allowing free users just 32K tokens compared to 128K for Plus users and a million for Pro and Ultra users.

To check your AI usage, click the cog icon in the Gemini app on the web or tap the menu button in the mobile app. The app will show you two bars: current usage which resets every five hours, and weekly limits that reset each week. If you hit these limits while on a paid plan, you’ll be demoted to the most basic AI model until the next reset.