GPT-6 Astra scores 70.5 on the Respan Index (#2), ahead of GPT-5.6 Terra at 53.5 (#14). GPT-6 Astra leads on 25 of the 26 benchmarks both report. GPT-5.6 Terra is 4.4x cheaper per token at list price.
* Estimated: no published scores in that category. How the Respan Index works
| 86% |
| 93.7% |
| Furniture Assembly (Epoch AI) | 54.2% | 80% |
| GPQA Diamond | 92.9% | 96% |
| LiveBench Agentic Coding (LiveBench) | 54.9% | 57.3% |
| LiveBench Coding (LiveBench) | 78.2% | 80.4% |
| LiveBench Data Analysis (LiveBench) | 79.3% | 83% |
| LiveBench Instruction Following (LiveBench) | 64.6% | 75.6% |
| LiveBench Language (LiveBench) | 82.9% | 89.4% |
| LiveBench Mathematics (LiveBench) | 94.9% | 96.8% |
| LiveBench Reasoning (LiveBench) | 90.6% | 92.7% |
| Mystery Game Puzzles (Epoch AI) | 35% | 84% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 99.7% | 100% |
| SAGE (Vals AI) | 47% | 46.4% |
| SimpleQA Verified | 43.2% | 75.6% |
| Terminal-Bench 4.0 (Vals AI) | 22.7% | 59.6% |
| Vending-Bench 2 (Andon Labs) | 7343.21 | 15514.7 |
| WeirdML (Håvard Tveit Ihle) | 78.3% | 93.6% |
| LMArena Agent (LMArena) | 0.004 | 0.1227 |
| LMArena Elo | 1466 | 1477 |
| LMArena Vision (LMArena) | 1268 | 1284 |
| BALROG (BALROG) | 53.2% | 68.3% |
| SimpleBench (SimpleBench) | 48.9% | 86.5% |
Higher is better. Each score links to where it was published.