Qwen3.8-Max scores 50.3 on the Respan Index (#22), ahead of Grok 4.7 at 49.8 (#23). Qwen3.8-Max leads on 13 of the 22 benchmarks both report. Both cost about the same per token at list price. Qwen3.8-Max has the larger context window (1M tokens).
* Estimated: no published scores in that category. How the Respan Index works
| FrontierSWE V2 (Proximal Labs) |
| 29.5% |
| 17.8% |
| Furniture Assembly (Epoch AI) | 20.8% | 20% |
| GPQA Diamond | 92.7% | 92.6% |
| LiveBench Agentic Coding (LiveBench) | 54% | 64.6% |
| LiveBench Coding (LiveBench) | 77.2% | 72.9% |
| LiveBench Data Analysis (LiveBench) | 76.9% | 78.4% |
| LiveBench Instruction Following (LiveBench) | 75.3% | 74.1% |
| LiveBench Language (LiveBench) | 80.1% | 79.7% |
| LiveBench Mathematics (LiveBench) | 95.7% | 91.3% |
| LiveBench Reasoning (LiveBench) | 82.7% | 88.2% |
| Mystery Game Puzzles (Epoch AI) | 29% | 38% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 98.1% | 99.4% |
| SAGE (Vals AI) | 40.8% | 51.3% |
| SimpleQA Verified | 56% | 45.8% |
| Terminal-Bench 4.0 (Vals AI) | 28.8% | 34.3% |
| LMArena Agent (LMArena) | 0.0404 | 0.0251 |
| LMArena Elo | 1443 | 1481 |
| LMArena WebDev (LMArena) | 1638 | 1671 |
Higher is better. Each score links to where it was published.