GPT-6 Astra scores 70.5 on the Respan Index (#2), ahead of Claude Sonnet 5.5 at 68.5 (#6). GPT-6 Astra leads on 17 of the 23 benchmarks both report. Claude Sonnet 5.5 is 5.0x cheaper per token at list price. GPT-6 Astra has the larger context window (1.1M tokens).
* Estimated: no published scores in that category. How the Respan Index works
| Furniture Assembly (Epoch AI) |
| 75% |
| 80% |
| GPQA Diamond | 95.6% | 96% |
| LiveBench Agentic Coding (LiveBench) | 56.3% | 57.3% |
| LiveBench Coding (LiveBench) | 91.4% | 80.4% |
| LiveBench Data Analysis (LiveBench) | 78.6% | 83% |
| LiveBench Instruction Following (LiveBench) | 70.5% | 75.6% |
| LiveBench Language (LiveBench) | 83.4% | 89.4% |
| LiveBench Mathematics (LiveBench) | 96.7% | 96.8% |
| LiveBench Reasoning (LiveBench) | 91.6% | 92.7% |
| Mystery Game Puzzles (Epoch AI) | 65% | 84% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 100% | 100% |
| SAGE (Vals AI) | 51.8% | 46.4% |
| SimpleQA Verified | 46.5% | 75.6% |
| Terminal-Bench 4.0 (Vals AI) | 64.1% | 59.6% |
| LMArena Agent (LMArena) | 0.1252 | 0.1227 |
| LMArena Elo | 1471 | 1477 |
| LMArena Vision (LMArena) | 1268 | 1284 |
| LMArena WebDev (LMArena) | 1786 | 1788 |
| SimpleBench (SimpleBench) | 75.9% | 86.5% |
Higher is better. Each score links to where it was published.