Reasoning Traces, Responses API, New SQLite Logs: LLM 0.32 Lands
LLM 0.32 is Simon Willison's command-line and Python tool for talking to language models, updated with visible reasoning traces, server-side provider tools, redesigned content-addressable SQLite logs, and support for the OpenAI Responses API.
What it is
The newest release of `llm`, Simon Willison's CLI and Python library for querying language models from many providers, described by its author as the most significant version since the project's initial launch.
What it does
Version 0.32 adds support for visible reasoning traces from thinking models, server-side provider tools instead of only client-executed tool calls, a redesigned content-addressable SQLite logging format, new model support, and features enabled by the OpenAI Responses API. A companion llm-anthropic 0.26 plugin release adds `claude-fable-5`, `claude-sonnet-5`, and `claude-opus-5` model support plus server-side WebSearch and WebFetch tools.
Why it matters
Server-side tools mean `llm` can hand off web search or fetch work to the provider instead of running it locally, and content-addressable SQLite logs make it easier to dedupe and query large histories of prompts and responses rather than re-fetching or re-storing repeated content.
How to use it
Upgrade with `pip install -U llm` or `uv tool upgrade llm`, then update the Anthropic plugin with `llm install -U llm-anthropic` to get the new Claude 5 model family and server-side tool support.