Mistral AI · released 2026-05-22 · modified MIT license (open-weight; self-hostable on as few as four GPUs) · announcement
unknown — not published, or the API model id isn't known yet.
256,000 tokens
not published as a table by Mistral AI.
no retirement announced.
no merged support PR yet on the runtimes AI Squawk watches (vLLM, llama.cpp, transformers, SGLang, Ollama), and not on the Ollama library.