Magistral: pricing, context window and benchmarks
Magistral is a multimodal model from Mistral AI. It offers a 200,000-token context window, costs $4.75 per million input tokens and $21.16 per million output tokens, and averages 79.8 across the index's benchmark suite.
Pricing
Magistral costs $4.75 per million input tokens and $21.16 per million output tokens. A blended 3:1 input-to-output mix works out to about $8.85 per million tokens. Cached input is listed at $0.44 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $4.75 | per 1M tokens |
| Output | $21.16 | per 1M tokens |
| Cached input | $0.44 | per 1M tokens |
| Blended 3:1 | $8.85 | estimate per 1M tokens |
| 1M in + 1M out | $25.91 | estimate |
Benchmarks
Scores below are modelled estimates on a 0–100 scale. Use them to compare relative position, not as official vendor results.
| Benchmark | Score | Rank | Percentile | Relative |
|---|---|---|---|---|
| MMLU-Pro | 82.1 | 120 | 80th | |
| GPQA Diamond | 75.8 | 126 | 79th | |
| SWE-bench Verified | 66.5 | 179 | 70th | |
| HumanEval | 85.1 | 111 | 82th | |
| MATH-500 | 85.4 | 101 | 83th | |
| AIME 2025 | 74.2 | 174 | 71th | |
| LiveCodeBench | 75.3 | 92 | 85th | |
| IFEval | 86.1 | 152 | 75th | |
| TruthfulQA | 82.0 | 129 | 79th | |
| HellaSwag | 87.9 | 146 | 76th | |
| MMMU | 77.4 | 109 | 82th | |
| Arena Elo | 1427.8 | 125 | 79th |
What Magistral is good for
Magistral is tuned toward coding work. It is a reasonable candidate for code generation, review, debugging and repository-scale tasks, especially when paired with a retrieval layer. At $4.75 per million input tokens, it sits in the mid-range of the index, and the 200,000-token context window is enough for large files and multi-file diffs.
Strengths
- Large 200,000-token context window for long documents and agent traces.
- Open weights make self-hosting, fine-tuning and data-residency review possible.
- Accepts more than text, including image input where the provider supports it.
- Coding benchmarks are a relative strength within its tier.
Trade-offs
- No standout weakness in the indexed specifications, but real-world behaviour still needs testing.
How it compares
| Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|
| Command A Turbo | Cohere | 32768 | $13.53 | 79.8 |
| ERNIE 4.5 Reasoning | Baidu | 1000000 | $14.36 | 79.8 |
| GLM-4.6 Chat | Zhipu AI | 400000 | $4.89 | 79.8 |
| Magistral | Mistral AI | 200000 | $4.75 | 79.8 |
| Qwen3 Instruct | Alibaba Cloud | 1000000 | $5.06 | 79.8 |
| Doubao 1.5 Pro | ByteDance | 200000 | $13.65 | 79.7 |
FAQ
How much does Magistral cost?
Magistral is listed at $4.75 per million input tokens and $21.16 per million output tokens. Cached input is $0.44 per million tokens. A blended 3:1 input-to-output mix is about $8.85 per million tokens.
What is the context window of Magistral?
Magistral has a context window of 200,000 tokens and can emit up to 128,000 output tokens per request in this index.
Is Magistral open weights?
Yes. Magistral is listed as an open-weight model under a open weights licence.
How does Magistral score on benchmarks?
Magistral averages 79.8 across the scored benchmarks in this index, ranking #126 of 600 models. Its strongest relative result is Arena Elo at 1427.8.
What is Magistral best for?
Magistral is tuned toward coding work. It is a reasonable candidate for code generation, review, debugging and repository-scale tasks, especially when paired with a retrieval layer. At $4.75 per million input tokens, it sits in the mid-range of the index, and the 200,000-token context window is enough for large files and multi-file diffs.