Claude Fable 5.1 (Medium): benchmark results
Provider: Anthropic. Released 2026-09-01. Access: API.
Aggregate Rating 1885 ± 19, rank #55 of 2170 rated models, from 35 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| RoboHarm | 20 | Safety refusal rate (%, 100 trials over 5 harmful instructio | 100 |
| Conceptual Reasoning Index | 74.95 | Chance-Corrected Score (0-100) | 97.1 |
| Conceptual Reasoning Index - Argument Evaluation (LMCA) | 65.32 | Chance-Corrected Score (0-100) | 96.1 |
| Conceptual Reasoning Index - Consistency (ACCoRD) | 83.67 | Chance-Corrected Score (0-100) | 95.7 |
| MedSafe-Dx v0.3 (4aj) | 71.1 | Balanced escalation utility (fixed-flag policy = 0, perfect | 94.7 |
| Conceptual Reasoning Index - Decision Theory (DTBench) | 95.12 | Chance-Corrected Score (0-100) | 93.4 |
| DuelLab GameBench 2 - Reasoning Variants | 82.92 | GameBench2 normalized score (0–100) | 92.5 |
| Epoch AI - Dtbench | 97.07 | Score | 92.5 |
| Epoch AI - Critpt | 29.14 | Score | 92.2 |
| ARC-AGI-2 | 86.25 | Accuracy (%) | 90.7 |
| ARC-AGI-1 | 94.5 | Accuracy (%) | 85.8 |
| Senior SWE-Bench | 34.74 | Tasteful Solve Rate (pass@1, %) | 84.1 |
Interactive version: theaggregate.ai/model?slug=claude-fable-5-1-medium · How It Works · Data refreshed daily, snapshot 2026-11-05.