You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You should limit the total output tokens (max_tokens + reasoning_tokens) to no more than 21,333 tokens (as the streaming API is required above that—this will soon be supported in Inspect but is not currently). Even with streaming, a maximum of 32,000 is recommended (above that they advise using batch mode).
The text was updated successfully, but these errors were encountered:
Just an issue to track this.
Docs say:
The text was updated successfully, but these errors were encountered: