Claude Sonnet 5 (Thinking) — benchmark results

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

Unified ELO 1858 ± 49, rank #83 of 1719 rated models, from 11 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Chatbot Arena (Code)1551Elo94.4
Chatbot Arena (Text)1464Elo91.5
Agent Arena - Confirmed Success12.24Confirmed Success (%)86.7
Chatbot Arena (Vision)1263Arena Score84.8
PM-LLM-Benchmark33.3Score74.8
Agent Arena - Praise vs Complaint13.43Praise vs Complaint (%)73.3
Chatbot Arena (Document)1476Elo67.7
Agent Arena - Tool Hallucination1.14Tool Hallucination (%)61.7
Agent Arena7.38Net Improvement (%)56.7
Agent Arena - Bash Recovery9.15Bash Recovery (%)53.3
Agent Arena - Steerability0.95Steerability (%)3.3

Interactive version: aibenchmarks.dev/model?slug=claude-sonnet-5-thinking · How the rankings work · Data refreshed daily, snapshot 2026-07-07.