Google fixes Gemini usage limits after single prompt maxed out subscriber's entire quota

Google Responds to Mounting Frustration Over Restrictive Usage Limits

Google has deployed several fixes to address growing user complaints about hitting limits too quickly with its AI assistant following the introduction of compute-based usage limits at I/O 2026. The changes mark a significant response to widespread criticism after the company switched from prompt-based quotas to a credit-style system that measures usage based on task complexity, features used, and conversation length 1

1

3

3

.

Josh Woodward, Vice President at Google, acknowledged in a post on X that users were encountering limits sooner than they should. The compute-used approach introduced five-hour refresh periods until weekly limits are met, but many subscribers found the new system far more restrictive than the previous experience 2

2

. One particularly striking incident involved a Google AI Pro subscriber who exhausted their entire five-hour AI usage cap after a single prompt using avatar-based Omni video generation, with the task running for three to four minutes before hitting 100% quota consumption and ultimately failing to produce any output 4

4

5

5

.

Major Bug Fixes Target Omni Video and Complex Prompts

One of the most significant fixes addresses a bug tied to Omni video generation that caused just one or two video prompts to consume disproportionate portions of user quotas. Someone experimenting with short clips or testing different styles could suddenly see their allowance drop far more than expected after only a couple of attempts 1

1

. Google has resolved this issue and is doubling the number of Omni video generations for Google AI Ultra subscribers immediately 3

3

.

Another area causing user complaints about hitting limits involved Gemini 3.1 Pro with complex prompts—long, detailed instructions often accompanied by large file uploads or multi-step reasoning tasks. These prompts were consuming quotas too aggressively. Google is now capping the amount of quota a single prompt can use, preventing extreme outliers where one task wipes out too much of a monthly allowance 1

1

2

2

.

Failed Requests No Longer Count Against Usage

In a change that addresses a particularly frustrating aspect of the user experience, Google confirmed that failed requests no longer count against usage quotas. Woodward noted that about one in ten requests can fail due to system errors, and previously even failed attempts could count against quotas 1

1

. Google clarified that "if a request fails, you won't be charged. Our system mistakes are on us, not you. Your quota is used only for successful completions" 3

3

.

Flash-Lite Becomes Free Tier as Google Adjusts Gemini's New Usage Limits

A notable update makes Gemini 3.1 Flash-Lite prompts completely free and exempt from quota consumption 1

1

2

2

. This effectively creates a free layer for lighter tasks and subtly encourages users to rely on lighter models when they don't need full reasoning power, helping stretch the limits of higher tiers further.

Google is also working on more detailed usage breakdowns and notifications for Deep Research and other compute-heavy tasks where Google Gemini processes large inputs or runs multi-step analysis. Many users currently have little visibility into why their quotas drop faster on some days than others 1

1

3

3

. The company aims to provide transparency so users can see which types of tasks are expensive and which are not.

Quality-of-Life Improvements and Future Plans

Google has implemented a useful improvement in how model selection works. Once you choose a specific model inside Google Gemini, the app will remember it across sessions. The selection will only change if you manually adjust it or hit a usage cap that triggers an automatic fallback to a lighter model 1

1

3

3

.

These changes represent Google's attempt to smooth out a system that had become inconsistent for many users. While Google has already increased Gemini usage limits for Antigravity users by as much as 9x compared to the immediate post-implementation period, broader caps for most regular subscribers appeared unchanged until these latest adjustments 4

4

. Looking ahead, Google plans to let users buy pay-as-you-go top-up AI credits to supplement their base allowances 3

3

. Whether these fixes fully resolve the frustration remains to be seen, but the direction now feels more transparent and user-friendly than the opaque system that sparked the initial backlash.🟡 untrained_description=🟡The image features the Gemini logo prominently on the left, next to a blue banner with a checkmark and the text "Limits Updated." On the right, a smartphone displays the Gemini application interface with the greeting "Hello, how can I help you today?" and options for "Deep Research," "Generate Image," and "Summarize Text." A close-up shot of a hand holding a smartphone with a blue case. The phone's screen shows a document titled "Optics Study Guide" within the Google Gemini application. The content visible on the screen discusses "Spherical Mirror Geometry & Equations" and "The Mirror Equation." At the bottom of the screen, the prompt "Ask Google Gemini" is visible, along with options for "Fast" input, a microphone icon, and a keyboard icon. The background is blurred, showing warm, out-of-focus lights. The image shows a person holding a red smartphone with the Google Gemini application open. The screen displays "What can I help with, Ab N.?" and indicates that "Gemini Flash" is the selected model. The Google 9TO5Google logo is visible at the bottom right of the image.🟡 output_schema=🟡{"summary": "string"}