Codestral: pricing, context window and benchmarks
Codestral is a multimodal model from Mistral AI. It offers a 65,536-token context window, costs $4.19 per million input tokens and $16.36 per million output tokens, and averages 77.6 across the index's benchmark suite.
Pricing
Codestral costs $4.19 per million input tokens and $16.36 per million output tokens. A blended 3:1 input-to-output mix works out to about $7.23 per million tokens. Cached input is listed at $0.42 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $4.19 | per 1M tokens |
| Output | $16.36 | per 1M tokens |
| Cached input | $0.42 | per 1M tokens |
| Blended 3:1 | $7.23 | estimate per 1M tokens |
| 1M in + 1M out | $20.55 | 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 | 79.2 | 203 | 66th | |
| GPQA Diamond | 75.3 | 138 | 77th | |
| SWE-bench Verified | 67.7 | 152 | 75th | |
| HumanEval | 80.5 | 218 | 64th | |
| MATH-500 | 82.1 | 161 | 73th | |
| AIME 2025 | 71.4 | 230 | 62th | |
| LiveCodeBench | 67.3 | 241 | 60th | |
| IFEval | 85.9 | 160 | 73th | |
| TruthfulQA | 79.9 | 179 | 70th | |
| HellaSwag | 88.2 | 132 | 78th | |
| MMMU | 76.0 | 139 | 77th | |
| Arena Elo | 1391.6 | 242 | 60th |
What Codestral is good for
Codestral 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 $4.19 per million input tokens, it sits in the mid-range of the index, and the 65,536-token context window is enough for large files and multi-file diffs.
Strengths
- 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 |
|---|---|---|---|---|
| GPT-5.1 Mini | OpenAI | 128000 | $12.19 | 77.7 |
| o4 Mini | OpenAI | 200000 | $11.92 | 77.7 |
| Codestral | Mistral AI | 65536 | $4.19 | 77.6 |
| Hunyuan-Turbo Pro | Tencent | 65536 | $12.00 | 77.6 |
| ERNIE 4.0 Pro | Baidu | 200000 | $12.58 | 77.5 |
| Hunyuan-Turbo Reasoning | Tencent | 128000 | $11.89 | 77.5 |
FAQ
How much does Codestral cost?
Codestral is listed at $4.19 per million input tokens and $16.36 per million output tokens. Cached input is $0.42 per million tokens. A blended 3:1 input-to-output mix is about $7.23 per million tokens.
What is the context window of Codestral?
Codestral has a context window of 65,536 tokens and can emit up to 4,096 output tokens per request in this index.
Is Codestral open weights?
Yes. Codestral is listed as an open-weight model under a open weights licence.
How does Codestral score on benchmarks?
Codestral averages 77.6 across the scored benchmarks in this index, ranking #183 of 600 models. Its strongest relative result is Arena Elo at 1391.6.
What is Codestral best for?
Codestral 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 $4.19 per million input tokens, it sits in the mid-range of the index, and the 65,536-token context window is enough for large files and multi-file diffs.