GPT-5.6 Terra scores 53.5 on the Respan Index (#14), ahead of Gemini 3.8 Flash at 52.2 (#16). Gemini 3.8 Flash leads on 14 of the 26 benchmarks both report. Gemini 3.8 Flash is 3.0x cheaper per token at list price. GPT-5.6 Terra 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) |
| 68.4% |
| 86% |
| Furniture Assembly (Epoch AI) | 31.7% | 54.2% |
| GPQA Diamond | 95.3% | 92.9% |
| LiveBench Agentic Coding (LiveBench) | 54.2% | 54.9% |
| LiveBench Coding (LiveBench) | 72.5% | 78.2% |
| LiveBench Data Analysis (LiveBench) | 54% | 79.3% |
| LiveBench Instruction Following (LiveBench) | 81.4% | 64.6% |
| LiveBench Language (LiveBench) | 87.8% | 82.9% |
| LiveBench Mathematics (LiveBench) | 91.6% | 94.9% |
| LiveBench Reasoning (LiveBench) | 89.3% | 90.6% |
| Mystery Game Puzzles (Epoch AI) | 47% | 35% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 98.9% | 99.7% |
| SAGE (Vals AI) | 35.1% | 47% |
| SimpleQA Verified | 69.7% | 43.2% |
| Terminal-Bench 4.0 (Vals AI) | 19.2% | 22.7% |
| Vending-Bench 2 (Andon Labs) | 5093.79 | 7343.21 |
| WeirdML (Håvard Tveit Ihle) | 84.8% | 78.3% |
| LMArena Agent (LMArena) | 0.0297 | 0.004 |
| LMArena Elo | 1495 | 1466 |
| LMArena Vision (LMArena) | 1290 | 1268 |
| SimpleBench (SimpleBench) | 82.4% | 48.9% |
| DeepSWE v1.1 | 73.7% | 69.6% |
Higher is better. Each score links to where it was published.