Claude Fable 5 (High): benchmark results
Claude Fable 5 evaluated at the high reasoning-effort setting. Provider: Anthropic. Released 2026-06-09. Access: API.
Unified ELO 1770 ± 1, rank #18 of 1761 rated models, from 32 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Deep20Bench | 12.06 | Questions to Solve (mean, lower is better) | 100 |
| EnigmaEval | 39.28 | Score (self-reported) | 100 |
| Epoch AI - Mirrorcode | 63.89 | Score | 100 |
| SEAL - EnigmaEval | 39.28 | Score | 100 |
| SWE-rebench | 64.5 | Resolved (%) | 100 |
| OTIS Mock AIME 2024-25 | 100 | Accuracy (%) | 99 |
| FutureEval | 13.77 | Unified Forecasting Score | 98.9 |
| Agent Arena - Steerability | 13.23 | Steerability (%) | 98.2 |
| Surface Evolver Bench | 95 | Mean Score (%) | 98 |
| Multi-turn Debate (Lechmazur) | 1747 | Bradley-Terry Rating | 97.8 |
| Agent Arena | 10.61 | Net Improvement (%) | 96.5 |
| Agent Arena - Praise vs Complaint | 20.25 | Praise vs Complaint (%) | 96.5 |
Interactive version: theaggregate.ai/model?slug=claude-fable-5-high · How It Works · Data refreshed daily, snapshot 2026-09-05.