llm-gemini Adds Gemini-3.5-Flash and Streaming

Simon Willison announced llm-gemini 0.32a0, a new plugin that integrates Google’s Gemini family of models into the llm framework, released 19 May 2026. The plugin requires llm>=0.32a0 alpha and introduces the ability to stream reasoning tokens, enabling finer-grained output as models generate chain-of-thought or intermediate steps. This matters for developers and researchers building applications that need real-time token-level reasoning or observability from Gemini models, and reflects ongoing ecosystem work to support major LLM providers within open ML tooling. The post links this release to other recent LLM developments and targets practitioners using the llm library and Google’s Gemini models.

Simon Willison released llm-gemini 0.32, an update to his LLM plugin that adds support for Google’s new gemini-3.5-flash model. The short post points readers to Willison’s separate notes on Gemini 3.5 Flash and mentions a pelican image generated using the updated plugin, highlighting practical experimentation. This matters for developers and researchers who use third-party integrators to access Gemini models, as plugin updates streamline adoption of new model variants and features. The release continues to track rapid LLM iteration and ecosystem tooling that connect open tooling to major model providers, affecting developer workflows and integration choices.