Claude Fable 5 (Low): benchmark results
Provider: Anthropic. Released 2026-06-09. Access: API.
Unified ELO 1725 ± 1, rank #72 of 3078 rated models, from 17 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OTIS Mock AIME 2024-25 | 97.78 | Accuracy (%) | 93 |
| Conceptual Reasoning Index - Argument Evaluation (LMCA) | 59.59 | Chance-Corrected Score (0-100) | 91.5 |
| InferenceBench | 8.74 | Speedup Score | 91.2 |
| Conceptual Reasoning Index | 68.69 | Chance-Corrected Score (0-100) | 88.6 |
| ARC-AGI-2 | 76.81 | Accuracy (%) | 87 |
| Conceptual Reasoning Index - Consistency (ACCoRD) | 78.92 | Chance-Corrected Score (0-100) | 86.4 |
| Opus Magnum Bench | 49.28 | Human-normalized score (%) | 85.7 |
| ObviousBench | 97.92 | Answer pass³ (%) | 80.9 |
| ReactBench | 41.2 | Pass@1 (%) | 79.6 |
| Conceptual Reasoning Index - Decision Theory (DTBench) | 85.78 | Chance-Corrected Score (0-100) | 77.6 |
| ARC-AGI-1 | 90.5 | Accuracy (%) | 77.3 |
| Chess Puzzles (Epoch AI) | 28 | Accuracy (%) | 76.4 |
Interactive version: theaggregate.ai/model?slug=claude-fable-5-low · How It Works · Data refreshed daily, snapshot 2026-09-20.