Claude Fable 5 (Max): benchmark results
Claude Fable 5 evaluated at the max reasoning-effort setting. Provider: Anthropic. Released 2026-06-09. Access: API.
Aggregate Rating 1904 ± 10, rank #33 of 2168 rated models, from 135 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| APEX-Agents-AA | 59.2 | Pass@1 (verified) | 100 |
| Composite-Bench | 74 | Macro Success (%) | 100 |
| HealthBench Professional (Mythos 5.1 System Card) | 63.3 | Length-Adjusted Score (%) | 100 |
| Kimi K3 Tech Report - Chat All-in-One Bench | 88 | Score (%) | 100 |
| Kimi K3 Tech Report - Kimi Code Bench 2.0 (Claude Code) | 76.9 | Score (%) | 100 |
| Kimi K3 Tech Report - MIRA Bench | 72.9 | Score (%) | 100 |
| Thinking Machines Inkling Launch - FORTRESS Adversarial | 96 | Harmful-Request Refusal Rate (%) | 100 |
| Toolathlon Pass@3 (Anthropic Internal Harness) | 68.5 | Pass@3 (%) | 100 |
| Vals AI LegalBench | 88.56 | Accuracy (%) | 100 |
| Vals AI Multimodal Index | 74.15 | Accuracy (%) | 100 |
| Vals AI Web Search | 48.45 | Accuracy (%) | 100 |
| Vals Index | 75.14 | Score (imported) | 100 |
Interactive version: theaggregate.ai/model?slug=claude-fable-5-max · How It Works · Data refreshed daily, snapshot 2026-10-20.