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 2104 ± 41, rank #8 of 1776 rated models, from 19 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Agent Arena - Praise vs Complaint23.94Praise vs Complaint (%)100
Creative Writing (Lechmazur)3.3Mean Score100
Multi-turn Debate (Lechmazur)1757.2Bradley-Terry Rating100
WeirdML87.85Average Score98.5
Surface Evolver Bench95Mean Score (%)97.7
Epoch AI - ECI160.75ECI Score96.5
PACT (Lechmazur)1603CMS Points96
Surface Evolver Bench Pass Rate87.5Pass Rate (%)95.5
Opus Magnum Bench60.15Human-normalized score (%)95.2
Agent Arena - Steerability14.62Steerability (%)94.6
Epoch AI - Cursorbench70.6Score92.6
Agent Arena12.72Net Improvement (%)91.9

Interactive version: theaggregate.ai/model?slug=claude-fable-5-high · How the rankings work · Data refreshed daily, snapshot 2026-07-22.