DeepSeek V4.1 Flash scores 50.9 on the Respan Index (#20), ahead of Gemini 3.7 Flash at 48.6 (#26). Gemini 3.7 Flash leads on 8 of the 15 benchmarks both report. DeepSeek V4.1 Flash is 2.9x cheaper per token at list price. Gemini 3.7 Flash has the larger context window (1.0M tokens).
* Estimated: no published scores in that category. How the Respan Index works
| LiveBench Instruction Following (LiveBench) | 70% | 79.9% |
| LiveBench Language (LiveBench) | 81.2% | 85.5% |
| LiveBench Mathematics (LiveBench) | 93.3% | 93.5% |
| LiveBench Reasoning (LiveBench) | 86.7% | 87.8% |
| SAGE (Vals AI) | 47.9% | 49.2% |
| Terminal-Bench 4.0 (Vals AI) | 19.7% | 12.1% |
| LMArena Agent (LMArena) | 0.0402 | -0.0149 |
| LMArena Elo | 1474 | 1488 |
| LMArena WebDev (LMArena) | 1620 | 1592 |
| DeepSWE v1.1 | 74.2% | 65.3% |
| GPQA Diamond | 90.9% | 94.8% |
Higher is better. Each score links to where it was published.