Mistral Large 2

About

Mistral Large 2 is Mistral AI's flagship open-weight model with 123 billion parameters, delivering exceptional multilingual performance across 12+ languages and strong competition with proprietary models like GPT-4o and Claude 3.5 Sonnet on reasoning and coding benchmarks. What distinguishes Mistral Large 2 from other open-weight models is its European language excellence — while most models treat French, German, Spanish, Italian, and Portuguese as secondary languages, Mistral was trained with native-level proficiency across these languages. The 128K token context window allows processing of substantial documents. Mistral Large 2 supports function calling and native JSON output, making it suitable for structured data extraction and agentic workflows. It's available under the Mistral Research License (open-weight for research and self-hosting) and through the Mistral platform API (USD 2.00 per 1M input tokens) with additional availability on Azure and AWS. The model's efficient architecture means it requires fewer computational resources than similarly sized competitors. For European companies requiring GDPR-compliant AI, developers building multilingual applications spanning European languages, and organizations wanting to self-host competitive AI without the resource requirements of larger models, Mistral Large 2 is an excellent choice. Compared to Llama 4, Mistral Large 2 offers better European language performance and more efficient architecture. The main limitations are text-only operation (no vision capabilities) and a smaller ecosystem compared to Meta's Llama models or OpenAI's API.

Best For

Pricing

Benchmarks

Technical Specs

Related Models