Simon Willison ran Mistral Large 4 through an SVG-generation prompt alongside Claude Opus 5.5, GPT-6.1 Sol, and Gemini 3.8 Flash, posting the results on October 6, 20261. The prompt, riffing on a Hacker News commenter's joke that "frontier models are tested with an armadillo in fishnet tights jaywalking on Mars," asked each model to generate exactly that scene as an SVG.
Willison used his open-source `llm` CLI tool to invoke each model at its default reasoning level, publishing rendered outputs via his Markdown SVG renderer. The four models tested were `claude-opus-5.5`, `gpt-6.1-sol`, `gemini-3.8-flash`, and `mistral/mistral-large-4`.
Mistral released Mistral Large 4 on October 6, a 1.05-trillion-parameter Mixture-of-Experts model that Mistral says will become the most capable open-weight system developed outside China once its weights ship on October 27[3]. Willison's post did not include qualitative rankings or benchmark scores for the SVG outputs.