Codestral Instruct: pricing, context window and benchmarks
Codestral Instruct is a multimodal model from Mistral AI. It offers a 65,536-token context window, costs $4.01 per million input tokens and $16.29 per million output tokens, and averages 75.1 across the index's benchmark suite.
Pricing
Codestral Instruct costs $4.01 per million input tokens and $16.29 per million output tokens. A blended 3:1 input-to-output mix works out to about $7.08 per million tokens. Cached input is listed at $0.58 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $4.01 | per 1M tokens |
| Output | $16.29 | per 1M tokens |
| Cached input | $0.58 | per 1M tokens |
| Blended 3:1 | $7.08 | estimate per 1M tokens |
| 1M in + 1M out | $20.30 | 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 | 78.4 | 228 | 62th | |
| GPQA Diamond | 69.2 | 287 | 52th | |
| SWE-bench Verified | 64.3 | 225 | 63th | |
| HumanEval | 82.5 | 171 | 72th | |
| MATH-500 | 78.1 | 256 | 57th | |
| AIME 2025 | 68.0 | 295 | 51th | |
| LiveCodeBench | 66.1 | 272 | 55th | |
| IFEval | 84.2 | 209 | 65th | |
| TruthfulQA | 78.2 | 248 | 59th | |
| HellaSwag | 87.3 | 161 | 73th | |
| MMMU | 70.0 | 302 | 50th | |
| Arena Elo | 1394.4 | 231 | 62th |
What Codestral Instruct is good for
Codestral Instruct is an open-weight model, so it can be self-hosted or run through a third-party inference provider. That matters for data residency, fine-tuning and long-run cost control. At $4.01 per million input tokens (hosted list price) and a 65,536-token context window, it is a practical option for teams that want more control than a closed API allows.
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.
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 |
|---|---|---|---|---|
| Codestral Instruct | Mistral AI | 65536 | $4.01 | 75.1 |
| Llama 3 Hyperbolic Chat | Hyperbolic | 200000 | $3.64 | 75.1 |
| Llama 3 Replicate Instruct | Replicate | 128000 | $3.71 | 75.1 |
| Mistral Small 3 Instruct | Mistral AI | 32768 | $3.97 | 75.1 |
| Reka Flash Turbo | Reka AI | 128000 | $10.19 | 75.1 |
| Nemotron 4 Chat | NVIDIA | 65536 | $3.99 | 75.0 |
FAQ
How much does Codestral Instruct cost?
Codestral Instruct is listed at $4.01 per million input tokens and $16.29 per million output tokens. Cached input is $0.58 per million tokens. A blended 3:1 input-to-output mix is about $7.08 per million tokens.
What is the context window of Codestral Instruct?
Codestral Instruct has a context window of 65,536 tokens and can emit up to 32,768 output tokens per request in this index.
Is Codestral Instruct open weights?
Yes. Codestral Instruct is listed as an open-weight model under a open weights licence.
How does Codestral Instruct score on benchmarks?
Codestral Instruct averages 75.1 across the scored benchmarks in this index, ranking #248 of 600 models. Its strongest relative result is Arena Elo at 1394.4.
What is Codestral Instruct best for?
Codestral Instruct is an open-weight model, so it can be self-hosted or run through a third-party inference provider. That matters for data residency, fine-tuning and long-run cost control. At $4.01 per million input tokens (hosted list price) and a 65,536-token context window, it is a practical option for teams that want more control than a closed API allows.