Mistral AI

Mistral AI was founded in Paris in April 2023 by Arthur Mensch, Guillaume Lample, and Timothée Lacroix, all veterans of DeepMind and Meta's AI research groups. The company built a reputation quickly by releasing its first model, Mistral 7B, as a free download that outperformed models twice its size, and followed with Mixtral 8x7B, one of the first prominent open-weight mixture-of-experts models. That open-first ethos, combined with strong benchmark performance relative to model size, made Mistral popular with developers who wanted capable models they could run themselves.

By January 2026, Mistral's annualized revenue had grown from roughly $20 million to $400 million in a single year, and the company raised $830 million in March 2026 to build data centers near Paris and in Sweden. As of May 2026, the flagship model lineup includes Mistral Large 3 (675B total parameters, 41B active, released December 2025), Mistral Small 4 (a unified MoE model combining reasoning, vision, and coding in one, released March 2026), and Mistral Medium 3.5 (frontier-class multimodal model optimized for agentic tasks, released April 2026). Codestral remains the company's coding-specialized model.

Mistral positions itself as Europe's answer to the US-dominated frontier AI landscape. Its open-weight models appeal to regulated industries, governments prioritizing digital sovereignty, and developers who want to self-host. The company's Le Chat application competes directly with ChatGPT and Claude.ai as a consumer assistant, and Le Chat Enterprise targets organizations building on Mistral's infrastructure. The Vibe coding tool and Mistral Forge (for enterprise custom model training) round out the product suite.