Grok 4.7 scores 49.8 on the Respan Index (#23), ahead of GLM-5.3 at 49.4 (#24). Grok 4.7 leads on 12 of the 22 benchmarks both report. GLM-5.3 is 1.4x cheaper per token at list price. GLM-5.3 has the larger context window (1M tokens).
* Estimated: no published scores in that category. How the Respan Index works
| FrontierSWE V2 (Proximal Labs) |
| 30.2% |
| 29.5% |
| GPQA Diamond | 90.9% | 92.7% |
| LiveBench Agentic Coding (LiveBench) | 60.9% | 54% |
| LiveBench Coding (LiveBench) | 79% | 77.2% |
| LiveBench Data Analysis (LiveBench) | 70.2% | 76.9% |
| LiveBench Instruction Following (LiveBench) | 69.3% | 75.3% |
| LiveBench Language (LiveBench) | 79.9% | 80.1% |
| LiveBench Mathematics (LiveBench) | 87.9% | 95.7% |
| LiveBench Reasoning (LiveBench) | 85.8% | 82.7% |
| Mystery Game Puzzles (Epoch AI) | 33% | 29% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 91.1% | 98.1% |
| SimpleQA Verified | 41% | 56% |
| Terminal-Bench 4.0 (Vals AI) | 38.9% | 28.8% |
| Vending-Bench 2 (Andon Labs) | 8163.61 | 10536.83 |
| LMArena Agent (LMArena) | 0.0236 | 0.0404 |
| LMArena Elo | 1479 | 1443 |
| LMArena WebDev (LMArena) | 1623 | 1638 |
| DeepSWE v1.1 | 66.9% | 71% |
Higher is better. Each score links to where it was published.