I’m experiencing a consistent issue in Google Antigravity IDE with Claude-powered agents (especially Sonnet 4.6 and Opus 4.6 thinking modes).
Problem:
The agent starts thinking normally, generation proceeds slowly, then it suddenly throws “The model’s generation exceeded the maximum output token limit” (or similar “maximum token output exceed”).
This happens even when the UI shows plenty of output tokens/context remaining and when generating long files/code that previously completed successfully in one pass.
The agent then retries automatically, often looping with the same error.
It used to handle truly long outputs reliably before recent updates.