Google fixes Gemini Omni bug, stops charging users for failed requests

Following complaints from Gemini users, Google has resolved the Omni video quota issue and introduced higher limits, free Flash-Lite prompts, and failed request protections

Google has fixed a bug in its Gemini Omni-powered video tool that caused usage quotas to be exhausted after just one or two requests, following user complaints about unexpectedly high consumption. The company said that failed AI requests will no longer count against user quotas, adding that usage will now be charged only for successful completions. It has also increased video generation limits for subscribers and introduced a set of changes aimed at making quota usage more predictable.

What happened with Google Gemini Omni

Google’s Gemini app switched from a daily quota limit to compute-based usage limits at Google I/O 2026. However, after this change, multiple users complained on social media platforms that they were hitting their usage limits much sooner than expected.

 

One user complained on X that a single attempt at generating an AI avatar video exhausted his entire five-hour usage allowance within minutes, despite the output reportedly failing to generate successfully. Google VP Josh Woodward took note of these complaints and said that he would look into them.

 

Now, Google has fixed that bug and announced a number of changes for Google AI plan subscribers.

  READ: Anthropic upgrades Claude AI with Opus 4.8: What's new in flagship model

What changes has Google made

Here’s an overview of all the changes that Google VP Woodward has announced:

• The bug causing just one or two Omni videos to drain quotas for certain users has been fixed.
• For Google AI Ultra members, Omni video generation limits have been doubled, effective immediately.
• Users will not be charged for failed requests. Woodward noted that system errors are on Google, not on users. Quota will now be used only for successful completions.
• Complex Gemini 3.1 Pro prompts, especially those involving large file attachments, were also depleting limits quickly. Google has now capped the amount of quota a single prompt can consume, allowing users to get more out of the Pro model.
• Flash-Lite prompts will be free and will not count against user quotas.
• Heavy tasks, such as Deep Research, require more compute. Google is designing more detailed usage breakdowns and notifications to help users better manage their limits.
• When a user selects a specific model, Google will remember that choice across sessions. It will only change if manually adjusted or if a usage cap triggers a fallback to a lighter model.