LLMOpen weights
Mistral Large 3
Mistral AI
Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error
Specifications
- Type
- LLM
- Access
- Open weights
- Released
- December 2, 2025
- License
- Apache 2.0
- Parameters
- 675B (41B active)
- Context window
- 262,144 tokens
- Output speed
- 54 tok/s
- Latency (TTFT)
- 0.64s
- Input
- Text, Image
- Output
- Text
- API pricing
- $0.50 in · $1.50 out / 1M tokens
Capabilities
Function callingStructured output
Benchmarks
Our take
Mistral’s flagship open MoE comeback (675B/41B) under Apache 2.0, debuting near the top of open non-reasoning leaderboards.