llm-gemini 0.34
Blog · Watch/Read · (gathered 2026-09-03)
- Simon Willison released
llm-gemini0.34, introducing support for Google’s newgemini-3.8-flashmodel, which offers low, medium, and high “thinking levels.” - The
gemini-3.8-flashmodel is highlighted for being “fast, cheap, and competent” at tasks like generating HTML and JavaScript, demonstrated by creating a “cool thing in HTML” in 13 seconds for 1.8 cents. - Willison also used Gemini 3.8 Flash with his
llm-coding-agentto add support for rendering HTML blocks in Markdown via sandboxed iframes within hismarkdown-svg-renderertool. - The update underscores the practical utility of Gemini Flash models for rapid, cost-effective code generation and their potential for integration into development workflows to enhance existing tools.