Google confirmed that a bug was responsible for this behavior. The company fixed the bug and clarified that failed AI requests will no longer count against user quotas, stating that usage would only be charged for successful completions going forward . In a further concession, Google also increased video generation limits for subscribers
.
Beyond the video bug, the structural change itself drew immediate criticism. Under the old system, users could send a predictable number of messages per day. Under the new system, heavier prompts and longer chats consume proportionally more quota, making limits feel unpredictable and often much tighter . Paid subscribers described the change as a "bait and switch" because they felt they were getting less usable access for the same monthly price
.
The frustration was especially acute for developers and creators doing intensive work; some reported triggering the 5-hour cooldown after only an hour or two of heavy use .
In response to what the company called "feedback about hitting limits too quickly," Google made several changes to the core Gemini app by late May :
The most aggressive adjustments were reserved for Google Antigravity, the company's agent-first coding platform. After paid users complained that they were hitting Gemini model rate limits too quickly inside Antigravity, Google tripled those rate limits—not once, but twice in a single week .
Antigravity lead Varun Mohan confirmed the platform reset everyone's weekly Gemini quota and increased rate limits across all paid tiers, including Gemini 3.5 Flash . The second increase specifically targeted weekly quotas, effectively giving paying Antigravity users a 9x boost compared to the initial post-I/O limits
.
Google also directed users of Gemini Code Assist IDE Extensions and Gemini CLI to migrate to Antigravity and Antigravity CLI before June 18, 2026, when those tools will stop serving requests for individual, AI Pro, and AI Ultra tiers .
Google now segments its Gemini access across several paid plans, with each tier multiplying the baseline limit. According to Google's support documentation :
| Plan | Price | Compute Limit Positioning |
|---|---|---|
| No plan | Free | Standard limits |
| AI Plus | $7.99/month | 2x higher than standard limits |
| AI Pro | $19.99/month | 4x higher than standard limits |
| AI Ultra | $100/month | 5x higher than AI Pro (up to 20x on specific plans) |
AI Ultra plans are designed for developers, technical leads, and advanced creators, and include priority access to Google Antigravity with the highest rate limits available . Higher-tier Ultra plans at $200/month offer the maximum 20x multiple over AI Pro
.
All plans remain subject to the five-hour refresh window and overarching weekly limit, meaning the multiplier gives you a larger bucket of compute, but heavy feature usage still drains that bucket faster than simple text prompts .
The overhaul arrived at a moment of extraordinary growth for Gemini. Google announced at I/O 2026 that the chatbot had reached 900 million monthly active users . The shift from message-based to compute-based quotas represented a structural necessity for a service now operating at massive scale.
At the same time, Google quietly prepared additional subscription tiers to smooth out the dramatic price and limit gaps between plans. Evidence of an upcoming "AI Ultra Lite" tier—positioned between the $20 Pro and $250 Ultra price points—surfaced in the Gemini macOS app, suggesting that the company recognized the jump between plans was too steep and intended to offer more granular access options .
The episode underscored a broader tension between the economics of serving frontier AI models and the expectations of subscribers who had grown accustomed to generous, predictable flat-rate access. In the days following the backlash, Google's adjustments were swift and targeted: fix the most egregious bugs, boost ceilings for the heaviest power users on Antigravity, and promise more flexible pay-as-you-go options down the line.
Comments
0 comments