Side-by-side comparison of DeepSeek Coder V2 (DeepSeek) and Devin (Cognition) — benchmarks, pricing, and capabilities.
| DeepSeek Coder V2 DeepSeek | Devin Cognition | |
|---|---|---|
| Category | Code Models | AI Agents |
| Specifications | ||
| Context Window | 128K | — |
| Pricing (per 1M tokens) | ||
| Input Cost | $0.14 | — |
| Output Cost | $0.28 | — |
| Performance | ||
| Overall Score | 84.5 | 84.0 |
| ARC-Challenge | 89.7 | — |
| GSM8K | 72.7 | — |
| HumanEval | 90.8 | — |
| MATH | 59.8 | — |
| MBPP | 86.5 | — |
| MMLU | 78.2 | — |
| SWE-bench Verified | — | 41.5 |
| TruthfulQA | 72.7 | — |
| WinoGrande | 85.9 | — |
| Community | ||
| User Rating | ★ 4.5 | ★ 4.1 |
| Reviews | 445 | 180 |