Google Gemini AI subscriptions quota bugs Gemini Ultra Google I/O

Google Admits Gemini Was Eating Your Quota Alive — and Finally Does Something About It

2 June 2026

Google has fixed several bugs in its Gemini app that were causing usage quotas to be consumed too quickly, such as one or two Omni videos depleting an entire quota and complex requests to the 3.1 Pro model with large files burning through too much allowance. As part of the fixes, Ultra members now receive double the Omni video generations, a cap has been placed on quota consumption per prompt, and failed requests are no longer charged. Additional improvements include free Flash Lite requests, more detailed consumption displays for complex features like Deep Research, and persistent model selection across sessions.

Google has quietly fixed a string of bugs in Gemini that were chewing through user quotas at an alarming rate. The admission came from Google VP Josh Woodward, who laid out the issues on X.

The worst offender: generating one or two Omni videos could wipe out your entire allocation in a single session. That bug is now patched. As a goodwill gesture, Ultra subscribers are getting double the Omni video generations going forward.

There was also a problem with complex prompts sent to the Gemini 3.1 Pro model, particularly those involving large files. These were consuming disproportionately large chunks of quota relative to what they should. Google has now capped the maximum quota consumption per prompt, meaning the request still runs normally — you just don't get punished for it.

A few other changes worth knowing about: failed requests are no longer billed against your quota, which is arguably the most basic fix imaginable but apparently needed doing. Flash Lite requests are now free entirely. Deep Research and other resource-heavy features will now show more granular usage breakdowns so you can actually see where your quota is going. And if you manually select a model, that choice will now persist across sessions rather than silently resetting itself.

None of this is glamorous. It's the kind of plumbing work that should have been sorted before charging people money. Still, better late than never.

The fixes arrive just days after Google's I/O event, where the company unveiled a redesigned Gemini app, an agent mode, and a revised subscription pricing structure. Whether the underlying quota system is now actually reliable remains to be seen.