Hi team,
I am currently subscribed to the Ultra plan and heavily rely on Antigravity IDE for my daily development. Recently, I’ve noticed an extreme spike in the quota refresh time for specific models.
While Gemini models refresh in a standard ~5 hours, the “Thinking” models (Claude Sonnet 4.6, Claude Opus 4.6) and GPT-OSS 120B are now showing a refresh time of 4 days and 22 hours.
Previously, the reset cycle for heavy models was around 3 hours. Jumping from 3 hours to nearly 5 days is a drastic change that essentially renders the Ultra plan unusable for continuous development.
Impact: Building autonomous agents and complex RAG pipelines for the Shopify ecosystem requires rapid, continuous testing and iteration. A 5-day lockout on logical reasoning models completely paralyzes my development workflow. It forces a hard stop on testing multi-model routing architectures.
Expected Behavior:
- A return to the previous ~3-hour refresh cycle, or at least a reasonable daily reset (e.g., 12-24 hours) for premium subscribers.
- Clearer communication or documentation if there has been a permanent policy change regarding rate limits for Anthropic/OSS models.
I have attached a screenshot of my settings panel for reference. Looking forward to a resolution or clarification on this matter.