GPT-5.5 scores 52.5 on the Respan Index (#15), ahead of Claude Opus 4.8 at 52.2 (#16). GPT-5.5 leads on 21 of the 32 benchmarks both report. Claude Opus 4.8 is 1.1x cheaper per token at list price.
* Estimated: no published scores in that category. How the Respan Index works
| 56.1% |
| 72.5% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 80% | 85.3% |
| Furniture Assembly (Epoch AI) | 42.5% | 44.2% |
| GPQA Diamond | 93.6% | 90.7% |
| LiveBench Agentic Coding (LiveBench) | 50.5% | 54% |
| LiveBench Coding (LiveBench) | 81.8% | 82.1% |
| LiveBench Data Analysis (LiveBench) | 66% | 81.6% |
| LiveBench Instruction Following (LiveBench) | 72% | 70.7% |
| LiveBench Language (LiveBench) | 79.7% | 87.4% |
| LiveBench Mathematics (LiveBench) | 94.3% | 95.9% |
| LiveBench Reasoning (LiveBench) | 89.2% | 89.7% |
| Mystery Game Puzzles (Epoch AI) | 36% | 56% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 98.3% | 84.4% |
| SAGE (Vals AI) | 54.8% | 51.5% |
| SimpleQA Verified | 53% | 63% |
| Vending-Bench 2 (Andon Labs) | 5787.43 | 7523.84 |
| WeirdML (Håvard Tveit Ihle) | 82.9% | 84.9% |
| LMArena Agent (LMArena) | 0.0664 | 0.0415 |
| LMArena Elo | 1453 | 1466 |
| LMArena Vision (LMArena) | 1286 | 1287 |
| AIME 2026 | 100% | 100% |
| tau2-bench Banking Knowledge (Sierra) | 39.7% | 44.6% |
| GSO Opt@1 (GSO) | 47.1% | 37.3% |
| Toolathlon-Verified (HKUST) | 76.2% | 73.5% |
| SimpleBench (SimpleBench) | 64.8% | 69% |
| BrowseComp | 84.3% | 84.4% |
| OSWorld-Verified | 83.4% | 78.7% |
| SWE-Bench Pro | 69.2% | 58.6% |
Higher is better. Each score links to where it was published.