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