Back to feed
Simon Willison·

llm 0.32a2

Signal
75
Hype
15
In three linesllm 0.32a2 adds support for OpenAI's `/v1/responses` endpoint for reasoning-capable models (GPT-5 class). Displays summarized reasoning tokens in distinct color. Use `-R` or `--hide-reasoning` flags to hide them.
Read source
Your take?
OpenAIReasoningTools

Summary generated by Claude — human-verified