Open-weight MoE with a 1M+ token context window and strong coding.
Major update with 10x context window expansion to over 1 million tokens
| Benchmark | DeepSeek V3.2 | DeepSeek-V3 | Δ |
|---|---|---|---|
| MMLU | 90.1% | 88.5% | +1.6 |
| HumanEval | 92.5% | 82.6% | +9.9 |
| Context Window | 1M+ tokens | — | — |
| MATH | 85.6% | 61.6% | +24.0 |
| GPQA | 68.4% | — | — |
| LiveCodeBench | 72.1% | — | — |