Google revises Gemini compute quotas after user complaints

Google has announced a series of adjustments to the new compute-based usage limits for its Gemini AI service following widespread user complaints that the caps were being exhausted too quickly. The changes were detailed by Gemini lead Josh Woodward in response to feedback from subscribers who reported hitting their five-hour or weekly quotas after just a handful of prompts, sometimes after a single, complex request.

The new system, introduced at the company's IO 2026 developer conference, replaced previous prompt-based limits with a model that tracks the computational resources each request consumes. Quotas reset every five hours up to a weekly maximum, with usage varying based on prompt complexity, tools used, and conversation length. However, users quickly reported issues, with some demonstrating that a single attempt to generate a video with Gemini Omni could drain an entire five-hour allowance before the task was complete.

In response, Google is implementing several fixes. A primary change caps the amount of quota any single prompt can consume, preventing one heavy request from depleting a user's entire session budget. The company has also clarified that failed requests or...