DeepSeek V4 Pro scores 46.9 on the Respan Index (#29), ahead of Qwen3.7-Max at 40.5 (#45). DeepSeek V4 Pro leads on 12 of the 18 benchmarks both report. DeepSeek V4 Pro is 1.9x cheaper per token at list price.
* Estimated: no published scores in that category. How the Respan Index works
| 54.9% |
| 43.6% |
| LiveBench Coding (LiveBench) | 77.2% | 74.2% |
| LiveBench Data Analysis (LiveBench) | 79.2% | 71.8% |
| LiveBench Instruction Following (LiveBench) | 67.7% | 74% |
| LiveBench Language (LiveBench) | 82.1% | 79.7% |
| LiveBench Mathematics (LiveBench) | 95.1% | 85.2% |
| LiveBench Reasoning (LiveBench) | 85.8% | 83.3% |
| Mystery Game Puzzles (Epoch AI) | 43% | 32% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 96.7% | 95.6% |
| SimpleQA Verified | 47% | 55.8% |
| SWE-bench Verified (Epoch AI) | 77.6% | 77.3% |
| LMArena Agent (LMArena) | 0.0119 | -0.0513 |
| LMArena WebDev (LMArena) | 1583 | 1515 |
| SimpleBench (SimpleBench) | 50.9% | 70.4% |
Higher is better. Each score links to where it was published.