Simon Willison Releases llm-gemini 0.32a0 — Adds Streaming Reasoning Tokens for Gemini Models

simonwillison.net·low signal

Simon Willison published llm-gemini 0.32a0 on May 19, a plugin update for his open-source LLM CLI tool that adds the ability to stream reasoning tokens from Google's Gemini models. Compatible with the llm>=0.32a0 alpha, this is a practical signal that reasoning/chain-of-thought transparency is becoming a standard feature across model providers and tooling. For builders using the LLM CLI, this means real-time visibility into Gemini's thinking process during inference.