Mistral AI

Mistral AI is a French AI company that has earned considerable respect in the AI industry for releasing high-quality language models at competitive cost and for its commitment to European AI sovereignty. Founded in 2023 by former DeepMind and Meta AI researchers, the company has moved quickly to release open-weight models that perform competitively with much larger proprietary alternatives.

The Mistral 7B model, released shortly after the company's founding, demonstrated that a model with 7 billion parameters could perform comparably to much larger models from well-funded competitors when trained and fine-tuned carefully. This efficiency has been a consistent theme in Mistral's model releases.

Mistral Large, the company's most capable proprietary model, competes with Claude and GPT-4 class models on benchmarks and is priced at $3/million input tokens, significantly cheaper than OpenAI and Anthropic's equivalents at comparable capability levels. Mistral Small and Mistral Nemo, cheaper models, provide strong performance for less demanding tasks at cost points below $0.30/million tokens.

The multilingual capability is a genuine differentiator for European and non-English language use cases. Mistral's models perform strongly across French, German, Spanish, Italian, and other European languages, which matters for organisations serving European markets.

Mistral has been vocal about European AI values, data handling within European jurisdiction, and maintaining open model weights, which has resonated with European organisations concerned about data sovereignty and dependence on US AI providers.

Le Chat, Mistral's consumer chat interface, is free and provides access to Mistral's models for everyday use. For developers, the Mistral API is cost-competitive and well-documented.