GPT-5.5 scores 52.5 on the Respan Index (#15), ahead of Gemini 3.8 Flash at 52.2 (#16). They split the 24 benchmarks both report evenly. Gemini 3.8 Flash is 7.5x cheaper per token at list price. Gemini 3.8 Flash has the larger context window (1.0M tokens).
* Estimated: no published scores in that category. How the Respan Index works
| 68.4% |
| 85.3% |
| Furniture Assembly (Epoch AI) | 31.7% | 44.2% |
| GPQA Diamond | 95.3% | 90.7% |
| LiveBench Agentic Coding (LiveBench) | 54.2% | 54% |
| LiveBench Coding (LiveBench) | 72.5% | 82.1% |
| LiveBench Data Analysis (LiveBench) | 54% | 81.6% |
| LiveBench Instruction Following (LiveBench) | 81.4% | 70.7% |
| LiveBench Language (LiveBench) | 87.8% | 87.4% |
| LiveBench Mathematics (LiveBench) | 91.6% | 95.9% |
| LiveBench Reasoning (LiveBench) | 89.3% | 89.7% |
| Mystery Game Puzzles (Epoch AI) | 47% | 56% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 98.9% | 84.4% |
| SAGE (Vals AI) | 35.1% | 51.5% |
| SimpleQA Verified | 69.7% | 63% |
| Vending-Bench 2 (Andon Labs) | 5093.79 | 7523.84 |
| WeirdML (Håvard Tveit Ihle) | 84.8% | 84.9% |
| LMArena Agent (LMArena) | 0.0297 | 0.0415 |
| LMArena Elo | 1495 | 1466 |
| LMArena Vision (LMArena) | 1290 | 1287 |
| SimpleBench (SimpleBench) | 82.4% | 69% |
Higher is better. Each score links to where it was published.