GPT-6 Astra scores 70.5 on the Respan Index (#2), ahead of Claude Fable 5.1 at 69.4 (#4). GPT-6 Astra leads on 16 of the 29 benchmarks both report. Both cost about the same 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
| FrontierMath Tiers 1-3 v2 (Epoch AI) |
| 90.2% |
| 93.7% |
| FrontierSWE V2 (Proximal Labs) | 56.3% | 65.5% |
| Furniture Assembly (Epoch AI) | 70% | 80% |
| LiveBench Agentic Coding (LiveBench) | 66.1% | 57.3% |
| LiveBench Coding (LiveBench) | 86.4% | 80.4% |
| LiveBench Data Analysis (LiveBench) | 80.3% | 83% |
| LiveBench Instruction Following (LiveBench) | 73% | 75.6% |
| LiveBench Language (LiveBench) | 89.5% | 89.4% |
| LiveBench Mathematics (LiveBench) | 97% | 96.8% |
| LiveBench Reasoning (LiveBench) | 91.7% | 92.7% |
| MirrorCode (Epoch AI) | 73.3% | 46.7% |
| Mystery Game Puzzles (Epoch AI) | 58% | 84% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 100% | 100% |
| SAGE (Vals AI) | 48.5% | 46.4% |
| SimpleQA Verified | 70.8% | 75.6% |
| Terminal-Bench 4.0 (Vals AI) | 58.1% | 59.6% |
| Vending-Bench 2 (Andon Labs) | 5421.56 | 15514.7 |
| WeirdML (Håvard Tveit Ihle) | 92.9% | 93.6% |
| LMArena Agent (LMArena) | 0.1431 | 0.1227 |
| LMArena Elo | 1501 | 1477 |
| LMArena Vision (LMArena) | 1288 | 1284 |
| LMArena WebDev (LMArena) | 1749 | 1788 |
| GSO Opt@1 (GSO) | 87.3% | 77.5% |
| SimpleBench (SimpleBench) | 86.6% | 86.5% |
| Terminal-Bench Science 0.1 | 52.6% | 64.6% |
Higher is better. Each score links to where it was published.