Curiosity · AI Model
Mistral Large 3
Mistral Large 3 is the European flagship from Mistral AI — an efficient, multilingual frontier model optimised for function calling, structured output, and European data-sovereignty requirements. It's the default for teams that need GDPR-friendly deployment without giving up reasoning quality.
Model specs
- Vendor
- Mistral AI
- Family
- Mistral Large
- Released
- 2025-11
- Context window
- 128,000 tokens
- Modalities
- text, code
- Input price
- $2/M tok
- Output price
- $6/M tok
- Pricing as of
- 2026-04-20
Strengths
- Excellent multilingual coverage beyond English
- Sovereign deployment options via Mistral La Plateforme and Azure EU regions
- Competitive price-performance for general-purpose work
Limitations
- Smaller ecosystem than OpenAI / Anthropic / Google
- Shorter context window than Gemini 2.5 Pro / Claude Opus 4.7
- No native vision or audio modalities
Use cases
- Multilingual assistants (especially French, German, Spanish)
- Structured-output pipelines requiring reliable JSON mode
- EU-hosted deployments under GDPR constraints
- Function-calling-heavy agent workflows
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| MMLU-Pro | ≈78% | 2026-04 |
| HumanEval | ≈85% | 2026-04 |
Frequently asked questions
What is Mistral Large 3?
Mistral Large 3 is a European frontier large language model from Mistral AI, released in late 2025. It is designed for multilingual applications, structured output, and GDPR-friendly deployment.
Where can I run Mistral Large 3 in Europe?
Through Mistral La Plateforme (France-hosted), Azure AI in EU regions, Vertex AI EU, and AWS Bedrock EU regions. It is also available on select sovereign-cloud deployments.
Sources
- Mistral AI — Models — accessed 2026-04-20