Claude Fable 5 scores 62.2 on the Respan Index (#9), ahead of Gemini 3.8 Flash at 52.2 (#16). Claude Fable 5 leads on 21 of the 27 benchmarks both report. Gemini 3.8 Flash is 13x 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
| FrontierMath Tiers 1-3 v2 (Epoch AI) |
| 87% |
| 68.4% |
| FrontierSWE V2 (Proximal Labs) | 47% | 19.6% |
| Furniture Assembly (Epoch AI) | 35.8% | 31.7% |
| GPQA Diamond | 85.9% | 95.3% |
| LiveBench Agentic Coding (LiveBench) | 62.2% | 54.2% |
| LiveBench Coding (LiveBench) | 86% | 72.5% |
| LiveBench Data Analysis (LiveBench) | 80.5% | 54% |
| LiveBench Instruction Following (LiveBench) | 75.8% | 81.4% |
| LiveBench Language (LiveBench) | 90.7% | 87.8% |
| LiveBench Mathematics (LiveBench) | 96% | 91.6% |
| LiveBench Reasoning (LiveBench) | 89.7% | 89.3% |
| Mystery Game Puzzles (Epoch AI) | 52% | 47% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 100% | 98.9% |
| SAGE (Vals AI) | 51.9% | 35.1% |
| SimpleQA Verified | 70.7% | 69.7% |
| Terminal-Bench 4.0 (Vals AI) | 41.4% | 19.2% |
| Vending-Bench 2 (Andon Labs) | 5680.26 | 5093.79 |
| WeirdML (Håvard Tveit Ihle) | 91.9% | 84.8% |
| LMArena Agent (LMArena) | 0.0821 | 0.0297 |
| LMArena Elo | 1504 | 1495 |
| LMArena Vision (LMArena) | 1309 | 1290 |
| LMArena WebDev (LMArena) | 1625 | 1583 |
| SimpleBench (SimpleBench) | 81.9% | 82.4% |
Higher is better. Each score links to where it was published.