Google has walked back its newly introduced compute-based usage limits for Gemini following backlash at its I/O event. The original caps drew sharp criticism from users who felt the restrictions were too aggressive.
Under the softened policy, failed requests no longer count against a user's quota, and prompts routed to Flash-Lite are now free. Ultra subscribers also get double the number of Omni generations.
Google additionally confirmed that pay-as-you-go credits are coming, giving heavy users a way to extend usage beyond their plan caps rather than being hard-blocked.
The reversal underscores how sensitive users have become to AI usage limits, and how quickly providers must respond when caps interfere with real workflows. Compute-based metering remains the underlying model, but Google is clearly tuning it to reduce friction.