Claude Fable 5: benchmark results
Anthropic Claude Fable 5, the generally available Claude 5-family frontier model for high-end reasoning, coding, and agentic tasks. Provider: Anthropic. Released 2026-06-09. Access: API.
Aggregate Rating 1907 ± 4, rank #16 of 1656 rated models, from 914 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA-Omniscience Index - Business | 34.5 | Omniscience Index | 100 |
| AGI-Eval Community - Cognition | 88.43 | Accuracy (%) | 100 |
| AGI-Eval Community - Cognition (Chinese) | 89.64 | Accuracy (%) | 100 |
| AGI-Eval Community - General Reasoning | 97.82 | Accuracy (%) | 100 |
| AGI-Eval Community - Mathematical Reasoning | 89.37 | Accuracy (%) | 100 |
| AIMultiple - HALC-Bench | 100 | Traps correctly answered 'not mentioned' (%, mean of three h | 100 |
| APEX-Agents-AA | 59.2 | Pass@1 (verified) | 100 |
| AgenticVBench (Claude Code) | 32.4 | Average Normalized Score (%) | 100 |
| Aiera Leaderboard - Grounding | 97.1 | Grounding axis (0-100) | 100 |
| Anthropic Single-Turn Harmful Requests - Harmless Rate (Claude.ai) | 99.54 | Harmless Response Rate (%) | 100 |
| Anthropic Suicide and Self-Harm Multi-Turn - Appropriate Response Rate (Claude.ai) | 100 | Appropriate Response Rate (%) | 100 |
| Appwrite Arena (Without Skills) | 97.7 | Overall Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=claude-fable-5 · How It Works · Data refreshed daily, snapshot 2026-10-20.