Mistral Large 3
Mistral Large 3 is a state-of-the-art general-purpose Multimodal granular Mixture-of-Experts model with 41B active parameters and 675B total parameters trained from scratch with 3000 H200s. This model is the instruct post-trained version, fine-tuned for instruction tasks, making it ideal for chat, agentic and instruction based use cases. Designed for reliability and long-context comprehension - It is engineered for production-grade assistants, retrieval-augmented systems, scientific workloads, and complex enterprise workflows.
Benchmark results
| Benchmark | Score | Tags | Source |
|---|---|---|---|
| Arena Hard | 55.1% | self-reported llm-stats | link → |
| MATH | 90.4% | self-reported llm-stats | link → |
| MATH (CoT) | 67.6% | self-reported llm-stats | link → |
| MM-MT-Bench | 84.9 | self-reported llm-stats | link → |
| MMLU-Redux | 82.0% | self-reported llm-stats | link → |
| MMMLU | 74.2% | self-reported llm-stats | link → |
| TriviaQA | 74.9% | self-reported llm-stats | link → |
| Wild Bench | 68.5% | self-reported llm-stats | link → |