GPT-5.6 Sol scores 60.6 on the Respan Index (#10), ahead of Grok 4.6 at 51.0 (#19). GPT-5.6 Sol leads on 24 of the 28 benchmarks both report. Grok 4.6 is 2.7x cheaper per token at list price. GPT-5.6 Sol has the larger context window (1.1M tokens).
* Estimated: no published scores in that category. How the Respan Index works
| 82.9% |
| 31.7% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 89.1% | 66% |
| FrontierSWE V2 (Proximal Labs) | 32.2% | 25.3% |
| Furniture Assembly (Epoch AI) | 56.7% | 40% |
| GPQA Diamond | 94.6% | 94% |
| LiveBench Agentic Coding (LiveBench) | 56.2% | 57% |
| LiveBench Coding (LiveBench) | 83.9% | 76.8% |
| LiveBench Data Analysis (LiveBench) | 79.8% | 73.9% |
| LiveBench Instruction Following (LiveBench) | 71.8% | 71.9% |
| LiveBench Language (LiveBench) | 87.7% | 83.7% |
| LiveBench Mathematics (LiveBench) | 96.2% | 92.6% |
| LiveBench Reasoning (LiveBench) | 91.7% | 90.5% |
| Mystery Game Puzzles (Epoch AI) | 58% | 34% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 100% | 99.2% |
| SAGE (Vals AI) | 52.6% | 28.9% |
| SimpleQA Verified | 69.7% | 49.3% |
| Terminal-Bench 4.0 (Vals AI) | 37.9% | 17.2% |
| Vending-Bench 2 (Andon Labs) | 9619.37 | 9047.03 |
| WeirdML (Håvard Tveit Ihle) | 89.4% | 67.3% |
| LMArena Agent (LMArena) | 0.0645 | 0.0128 |
| LMArena Elo | 1484 | 1454 |
| LMArena Vision (LMArena) | 1285 | 1264 |
| DeepSWE v1.1 | 72.7% | 65.9% |
| SimpleBench (SimpleBench) | 71.7% | 75.9% |
Higher is better. Each score links to where it was published.