The Gemini users have been complaining for long about the Gemini access running out too quickly. Google has now responded to the complaints, as the Mountain View-based tech giant is now making fresh changes to how usage limits work inside the Gemini app. Moreover, the firm also recently announced during their Google I/O 2026 that they are shifting to a compute-based system. Following that, Google has now said that they are adding safeguards so users can get more value from Gemini 3.1 Pro without losing large parts of their quota in a single request. The update also brings better tracking tools, free access to Flash-Lite prompts, and fixes for a bug that caused some video tasks to consume unusually high amounts of quota for many users across regions globally.
According to Google, if someone gives lighter text prompts, then it uses far less computing power than advanced requests involving coding, long conversations, or video generation. The company also added that this approach is designed to better match how much work the AI actually performs. Reports also suggest that the new system refreshes every five hours until a weekly limit is reached.
Gemini lead Josh Woodward said Google is now limiting how much quota one prompt can consume while using Gemini 3.1 Pro. He said Google is taking the step following the users’ reports where they said that large files and complex prompts were draining their limits too quickly. Moreover, the tech giant also clarified that failed requests will not reduce a user’s available quota. The company said users will only be charged for successful completions, while system errors will not count against limits.
At the same time, Google plans to introduce more detailed usage reports for heavy tools such as Deep Research. The current usage dashboard only gives users a broad overview of their remaining quota. Future updates will include clearer breakdowns and notifications to help people manage usage better.
Also read: Acer unveils Predator Atlas 8 gaming handheld ahead of Computex 2026: Features and top specs
Another key change is that the Gemini 3.1 Flash-Lite prompts are now completely free and will not count toward usage limits. Google also confirmed that the app will remember a user’s selected AI model across future sessions unless the user manually changes it or hits a limit that triggers a fallback to a lighter model.
The company additionally fixed an issue where generating just one or two Omni videos was rapidly exhausting quotas for some users. Google AI Ultra subscribers will now receive double the number of Omni video generations.