Google has announced changes to the limit system for users of its Gemini AI assistant. The decision was in response to numerous complaints from the community that they were exhausting their allocated limits too quickly after the implementation of a new request scoring system.
Google has announced changes to the limit system for users of its Gemini AI assistant. The decision was in response to numerous complaints from the community that they were exhausting their allocated limits too quickly after the implementation of a new request scoring system.
The innovation was announced by Gemini head Josh Woodward, 9to5Google reports. In particular, the tech giant plans to allow the purchase of additional AI credits based on actual usage, as well as provide users with more detailed analytics on power consumption during complex tasks such as Deep Research.
Last week, Google switched to a new approach to limits, which are updated every 5 hours until a weekly quota is reached. This system calculates the load depending on the complexity of the queries, the tools involved, and the duration of the dialogue. Simple text consumes minimal resources, while working with code or video requires significant power. Because of this, users who uploaded large files to Gemini 3.1 Pro instantly lost access to the model.
To fix this, Google will artificially limit the amount of quota that a single super-complex query can consume. According to Woodward, this is done to «limit the amount of quota that a single query can use so that you can get more out of the Pro model.»
The company also clarified that system errors will not affect users' balances. «If a request fails, you will not be charged. Errors in our system are our responsibility, not yours. Your limit is used only for successful completions,» Google notes.
Another important change concerns the 3.1 Flash-Lite model: requests to it are now completely free and do not count towards the overall quota. The system will automatically remember the model selection for future sessions, and the transition to the lighter version will occur only in case of exhaustion of limits or manual configuration.
Additionally, we’ve fixed an issue that caused some users to prematurely reach their Omni multimodal video processing limits. Google AI Ultra subscribers will now receive twice as many Omni generations.
