DeepSeek V4 Pro — Benchmarks
Benchmark scores for DeepSeek V4 Pro aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.
Overall rank: #30 of 123 open modelscomposite 59/100 across 9 benchmarks in 4 categories · methodology
Coding
| Benchmark | Score | Open rank | All models |
|---|---|---|---|
| SciCode | 50.0% | #6 / 50 | #58 / 155 |
| LiveBench Coding | 70.0 | #14 / 16 | #48 / 54 |
Knowledge
| Benchmark | Score | Open rank | All models |
|---|---|---|---|
| SimpleQA | 47.0% | #3 / 15 | #28 / 78 |
Math
| Benchmark | Score | Open rank | All models |
|---|---|---|---|
| AIME 2024/2025 | 96.7% | #3 / 74 | #24 / 270 |
| LiveBench Math | 90.7 | #2 / 16 | #23 / 54 |
| ProofBench | 16.0% | #10 / 21 | #41 / 64 |
Reasoning
| Benchmark | Score | Open rank | All models |
|---|---|---|---|
| GPQA Diamond | 90.9% | #5 / 83 | #27 / 291 |
| LiveBench Reasoning | 82.7 | #7 / 16 | #35 / 54 |
| CritPt | 12.9% | #7 / 50 | #57 / 167 |
Scores aggregated from public benchmark sources (each linked from the benchmark pages). llmrun does not run these benchmarks.