Mistral AI
European AI lab with strong open-weight models. Mistral Large 3 is the flagship — 675B MoE (41B active), Apache 2.0 license, ~1418 ELO on LMArena, 256k context, multimodal. #2 among open-source non-reasoning models. Codestral 25.08 for production code generation (80+ languages). Mistral Medium 3 ($0.40/$2) for efficient tasks. Voxtral Transcribe 2 for speech (4B params, runs on-device). EU-based with strong data sovereignty. Pricing: Medium ~$0.40/$2, Large ~$2/$6 per 1M tokens.
Overview
| Category | Ai |
| Compliance | SOC2, GDPR |
| Self-Hostable | Yes |
| On-Prem | No |
| Best For | startup, growth, enterprise |
| Last Verified | 2026-02-12 |
Strengths & Weaknesses
Strengths:- cost
- performance
- security
- Smaller ecosystem and fewer integrations than OpenAI
- Documentation less mature
- Not competitive with top-3 frontier models (Opus 4.6, Gemini 3 Pro, Grok 4.1)
When to Use
Best when:- EU data residency is required
- Want open-weight models (Apache 2.0) with API fallback
- Need strong code generation (Codestral 25.08)
- Need on-device speech transcription (Voxtral)
- Cost-sensitive with 256k context needs
- Need absolute frontier quality (#1-3 on LMArena)
- Require maximum ecosystem compatibility