Skip to content

docs(gpt-6): tool-call effort is honoured up to high - #133

Merged
VickyXAI merged 1 commit into
mainfrom
docs/gpt-6-effort-cap
Oct 10, 2026
Merged

VickyXAI merged 1 commit into
mainfrom
docs/gpt-6-effort-cap

Conversation

@VickyXAI

Copy link
Copy Markdown
Contributor

This matches the docs to blockrun #811. GPT-6 tool calls run at the caller's reasoning_effort up to high; xhigh and max are served at high.

Why: the per-call price does not yet account for effort. Measured on gpt-6.1-sol, a tool call at xhigh used about 3x the completion tokens of the same call at low, for the same charge.

#130 said max on tool calls is sent as asked. That is no longer what ships.

The edits are prose only and no counts change.

🤖 Generated with Claude Code

The gateway caps GPT-6 tool calls at reasoning_effort "high" (blockrun #811):
the per-call price does not yet account for effort, and xhigh measured ~3x
the completion tokens of low at the same charge. #130 said max on tool calls
was sent as asked; that is no longer what ships.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
@VickyXAI
VickyXAI merged commit 81db787 into main Oct 10, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant