GPT-6 Astra scores 70.5 on the Respan Index (#2), ahead of GPT-5.6 Sol at 60.6 (#10). GPT-6 Astra leads on 25 of the 30 benchmarks both report. GPT-5.6 Sol is 2.5x cheaper per token at list price.
* Estimated: no published scores in that category. How the Respan Index works
| 82.9% |
| 97.6% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 89.1% | 93.7% |
| FrontierSWE V2 (Proximal Labs) | 32.2% | 65.5% |
| Furniture Assembly (Epoch AI) | 56.7% | 80% |
| GPQA Diamond | 94.6% | 96% |
| LiveBench Agentic Coding (LiveBench) | 56.2% | 57.3% |
| LiveBench Coding (LiveBench) | 83.9% | 80.4% |
| LiveBench Data Analysis (LiveBench) | 79.8% | 83% |
| LiveBench Instruction Following (LiveBench) | 71.8% | 75.6% |
| LiveBench Language (LiveBench) | 87.7% | 89.4% |
| LiveBench Mathematics (LiveBench) | 96.2% | 96.8% |
| LiveBench Reasoning (LiveBench) | 91.7% | 92.7% |
| MirrorCode (Epoch AI) | 20% | 46.7% |
| Mystery Game Puzzles (Epoch AI) | 58% | 84% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 100% | 100% |
| SAGE (Vals AI) | 52.6% | 46.4% |
| SimpleQA Verified | 69.7% | 75.6% |
| Terminal-Bench 4.0 (Vals AI) | 37.9% | 59.6% |
| Vending-Bench 2 (Andon Labs) | 9619.37 | 15514.7 |
| WeirdML (Håvard Tveit Ihle) | 89.4% | 93.6% |
| LMArena Agent (LMArena) | 0.0645 | 0.1227 |
| LMArena Elo | 1484 | 1477 |
| LMArena Vision (LMArena) | 1285 | 1284 |
| GSO Opt@1 (GSO) | 70.6% | 77.5% |
| BALROG (BALROG) | 60% | 68.3% |
| SimpleBench (SimpleBench) | 71.7% | 86.5% |
Higher is better. Each score links to where it was published.