Claude Sonnet 5 (Non-reasoning): benchmark results

Provider: Anthropic. Released 2026-06-30. Access: API.

Aggregate Rating 1701 ± 26, rank #353 of 2151 rated models, from 13 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Shade Coding Prompt Injection - Attack Success Rate (Mythos 5.1 System Card)0.15Attack Success Rate (%)100
LinguaGuard — original wording balanced accuracy — structured step prompt0.96Balanced accuracy92.3
Shade Coding Prompt Injection - Attack Success Rate with Probes (Mythos 5.1 System Card)0.2Attack Success Rate (%)80
Shade Coding Prompt Injection, Stronger Attacker - Attack Success Rate (Mythos 5.1 System Card)40.3Attack Success Rate (%)80
Shade Computer Use Prompt Injection - Attack Success Rate with Probes (Mythos 5.1 System Card)1.04Attack Success Rate (%)60
DuelLab Overall28.43GameBench2 normalized score (0–100)38.9
BBQ Ambiguous Accuracy (Sonnet 5 & Opus 5 System Cards)98.6Accuracy (%)14.3
BBQ Disambiguated Accuracy (Mythos 5.1 System Card)72.36Accuracy (%)0
Shade Coding Prompt Injection, Stronger Attacker - Attack Success Rate with Probes (Mythos 5.1 System Card)37.51Attack Success Rate (%)0
Shade Computer Use Prompt Injection - Attack Success Rate (Mythos 5.1 System Card)4.86Attack Success Rate (%)0
StateSight - Connected Components7.3Accuracy (%; counting 4-neighbor connected gray regions in a0
StateSight - Cube Net Folding53.3Accuracy (%; naming the face opposite a highlighted face aft0

Interactive version: theaggregate.ai/model?slug=claude-sonnet-5-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-17.