llm-gemini 0.34 adds Gemini 3.8 Flash with three thinking levels the day after Google shipped it
Simon Willison's Weblog·low signal
Simon Willison released llm-gemini 0.34 on 2026-09-02, adding the gemini-3.8-flash model with low, medium and high thinking levels (issue #146) and fixing async responses that were failing to record the resolved model version. The async model-version bug is the more useful detail for anyone logging LLM calls, since it means prior async runs recorded through this plugin cannot be attributed to a specific model build.