How Google Gemini new compute limits affect heavy AI users

Unlike ChatGPT, Claude, and many other AI platforms, Gemini has historically employed a daily prompt limit that was more generous than its competitors.

However, at I/O 2026, Google announced a shift from this prompt-based system to a “compute-used” model. Previously, Gemini’s daily limit was based on prompt count, differing from ChatGPT and Claude, which based limits on token usage.

Moving forward, Gemini will measure limits based on prompt complexity, features used, and chat length—richer tasks like video generation will consume more compute than simple text.

For basic use, this change shouldn’t impact your workflow, but heavy users performing complex tasks such as video creation or coding might reach limits sooner. Similar to other services, Gemini will now have five-hour and weekly compute limits. Exceeding compute usage will first hit the five-hour cap, then the weekly limit.

Details are available on Google’s support page. Without a plan, default “standard limits” apply; AI Plus doubles those limits, AI Pro quadruples them, and AI Ultra offers 20 times higher usage than AI Pro.

Activities like specific prompts and actions will also count toward your limit. You can monitor your usage via the Gemini app or web. Once limits are reached, Google switches you to a smaller model, but AI Pro and Ultra subscribers can buy PAYG credits to extend usage, which will be available soon in Google Antigravity, Google Flow, and the Gemini app.

These updates coincide with the launch of a new $100 AI Ultra plan and a $50 price cut for the top-tier plan, now costing $200 per month.