Claude Sonnet 4.6 (Medium) — benchmark results

Provider: Anthropic. Released 2026-02-17. Access: API.

Unified ELO 1800 ± 33, rank #163 of 1839 rated models, from 8 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
ALE-Bench1327.3Performance (Self-Refine x1) (self-reported)94.9
WeirdML66.07Average Score81.3
Epoch AI - ECI152.35ECI Score74.6
OTIS Mock AIME 2024-2582.22Accuracy (%)65.7
GIM0.84IRT ability (theta)64.4
Opus Magnum Bench12.37Human-normalized score (%)23.8
Chess Puzzles (Epoch AI)8Accuracy (%)16.9
Epoch AI - Cursorbench46Score3.3

Interactive version: theaggregate.ai/model?slug=claude-sonnet-4-6-medium · How It Works · Data refreshed daily, snapshot 2026-08-05.