I checked the server logs for your request (7fc406e8-...). The agent was actually actively working throughout the session.
What you saw as “stuck at thinking” is Opus 4.7’s extended thinking behavior. This model can think for 15-60+ seconds between actions, which looks like a freeze in the UI even though it’s still processing. The “10M+ tokens” displayed likely reflect thinking tokens (internal reasoning), not actual usage.
A few tips if you run into this again:
Press Stop and resend — this often recovers the conversation if it feels genuinely stuck
Start a new chat for complex multi-step tasks rather than continuing very long sessions
Try a different model (like Sonnet 4.5) if you want faster response times — Opus trades speed for deeper reasoning
Our team is aware this UX can be confusing and is investigating improvements to how extended thinking progress is displayed.
but the UI has not been updated after a day (still display thinking), and i did not get the result but cost 10M+ tokens, can u help me cancel and return this cost for me?
