Google eases Gemini usage restrictions following user feedback.

At last week's Google I/O conference, Google announced that Gemini, which now has 900 million monthly active users, will be transitioning from a daily query limit to a usage limit based on computing power.

These limits will be refreshed every 5 hours until the user reaches their weekly limit. However, because the calculation depends on the complexity of the statement and the resources used, many users have encountered difficulties as their limits are consumed too quickly.

Google Gemini is currently the world's leading chatbot.

To address this issue, Josh Woodward, Vice President of Google Labs, Gemini, and AI Studio, said the company is implementing several solutions to help extend user limits and make them more predictable. One of the key changes is that commands using Google's Flash-Lite model "should be free and won't count toward your limit."

Furthermore, a user's spending limit will not be affected if a request fails to process. Explaining this, Mr. Woodward stated: "System errors are our responsibility, not yours. Your spending limit is only used for successful requests."

For users who utilize the resource-intensive Research mode, Gemini will now display more detailed usage notifications and analytics. This aims to help users proactively manage their usage and avoid exceeding their limits too quickly. Additionally, for complex commands using the Gemini 3.1 Pro model, especially those with large file attachments, Google will also apply a cap to the amount of resources a single command can consume.

Google also confirmed it has fixed a bug that caused videos created using Omni – the company's new world model capable of generating content from various input types – to consume more usage limits than expected for some users.

To compensate, the company decided to double the number of Omni video creations for customers using the Google AI Ultra plan.

Finally, a new and useful feature is that Gemini will remember the last model the user selected for all future sessions. The system "will only change if you manually adjust it or when a certain limit is reached, causing it to automatically revert to a lighter model."

This new move shows Google is seeking to strike a balance between managing the computational costs of advanced AI models and the user experience, amid increasingly fierce competition in the artificial intelligence market.