GPT-6 Astra scores 70.5 on the Respan Index (#2), ahead of GPT-6 Sol at 62.8 (#8). GPT-6 Astra leads on 22 of the 24 benchmarks both report. GPT-6 Sol is 5.0x cheaper per token at list price.
* Estimated: no published scores in that category. How the Respan Index works
| Furniture Assembly (Epoch AI) |
| 80% |
| 58.3% |
| GPQA Diamond | 96% | 94.3% |
| LiveBench Agentic Coding (LiveBench) | 57.3% | 52.9% |
| LiveBench Coding (LiveBench) | 80.4% | 81.8% |
| LiveBench Data Analysis (LiveBench) | 83% | 81.2% |
| LiveBench Instruction Following (LiveBench) | 75.6% | 68.6% |
| LiveBench Language (LiveBench) | 89.4% | 85.3% |
| LiveBench Mathematics (LiveBench) | 96.8% | 96.4% |
| LiveBench Reasoning (LiveBench) | 92.7% | 88.7% |
| Mystery Game Puzzles (Epoch AI) | 84% | 56% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 100% | 100% |
| SAGE (Vals AI) | 46.4% | 44.8% |
| SimpleQA Verified | 75.6% | 60.7% |
| Terminal-Bench 4.0 (Vals AI) | 59.6% | 44.4% |
| Vending-Bench 2 (Andon Labs) | 15514.7 | 14427.85 |
| LMArena Agent (LMArena) | 0.1227 | 0.0971 |
| LMArena Elo | 1477 | 1457 |
| LMArena WebDev (LMArena) | 1788 | 1689 |
| SimpleBench (SimpleBench) | 86.5% | 73.1% |
| AutomationBench 1.0.6 | 41.4% | 33.2% |
Higher is better. Each score links to where it was published.