Side-by-side comparison of Codestral (Mistral AI) and DeepSeek V3 (DeepSeek) — benchmarks, pricing, and capabilities.
| Codestral Mistral AI | DeepSeek V3 DeepSeek | |
|---|---|---|
| Category | Code Models | LLMs |
| Specifications | ||
| Context Window | 32K | 128K |
| Pricing (per 1M tokens) | ||
| Input Cost | $0.20 | $0.27 |
| Output Cost | $0.60 | $1.10 |
| Performance | ||
| Overall Score | 86.2 | 90.0 |
| BigBench Hard | — | 85.5 |
| DROP | — | 85.0 |
| GSM8K | 84.0 | 93.0 |
| HumanEval | 92.5 | 89.2 |
| MATH | 60.5 | 75.8 |
| MBPP | 88.0 | — |
| MMLU | 79.0 | 87.1 |
| TruthfulQA | — | 68.0 |
| Community | ||
| User Rating | ★ 4.6 | ★ 4.5 |
| Reviews | 289 | 620 |