Home Model series Claude Sonnet Claude Sonnet 5 Claude Sonnet 5 · Toolathlon · June 2026 release used in selected source · Scores and configurations Claude Sonnet 5 · Toolathlon · June 2026 release used in selected source · Scores and configurations Claude Sonnet 5 record for Toolathlon · June 2026 release used in selected source. 4 score records for this release, including 4 exact values (may include historical records). Includes 1 source location.
Scores and evaluation conditions Claude Sonnet 5 · Toolathlon scores and configurations Metric and score Configuration and evaluation conditions Sources Pass@1 54.3% Reported 54.3% · Unit % Published Unconfirmed
Standard configuration: adaptive thinking, max effort Settings Adaptive thinking at max effort; default temperature and top_p; five-trial averages unless otherwise stated. Unknowns Context-window size is evaluation-dependent; standard configuration is 1M tokens. Tools Tool access not recorded Reasoning budget max Framework not recorded Protocol Toolathlon 108 tasks, three trials, internal harness. Conditions Adaptive thinking and max effort; offline snapshots pin financial feeds and container images; about a quarter of tasks are unsatisfiable as published. View this score in rankings Pass@3 63% Reported 63.0% · Unit % Published Unconfirmed
Standard configuration: adaptive thinking, max effort Settings Adaptive thinking at max effort; default temperature and top_p; five-trial averages unless otherwise stated. Unknowns Context-window size is evaluation-dependent; standard configuration is 1M tokens. Tools Tool access not recorded Reasoning budget max Framework not recorded Protocol Toolathlon 108 tasks; at least one correct result across three trials. Conditions Adaptive thinking and max effort; internal harness with pinned dependencies. View this score in rankings Pass³ 40.7% Reported 40.7% · Unit % Published Unconfirmed
Standard configuration: adaptive thinking, max effort Settings Adaptive thinking at max effort; default temperature and top_p; five-trial averages unless otherwise stated. Unknowns Context-window size is evaluation-dependent; standard configuration is 1M tokens. Tools Tool access not recorded Reasoning budget max Framework not recorded Protocol Toolathlon 108 tasks; all three attempts correct. Conditions Adaptive thinking and max effort; internal harness with pinned dependencies. View this score in rankings Average assistant turns 26 turns Reported 26.0 · Unit turns Published Unconfirmed
Standard configuration: adaptive thinking, max effort Settings Adaptive thinking at max effort; default temperature and top_p; five-trial averages unless otherwise stated. Unknowns Context-window size is evaluation-dependent; standard configuration is 1M tokens. Tools Tool access not recorded Reasoning budget max Framework not recorded Protocol Average assistant turns per trajectory on Toolathlon. Conditions Adaptive thinking and max effort; three trials per task. View this score in rankings
Benchmark versions and metrics Aliases Toolathlon
Pass@1 · % · Higher is better Pass@3 · % · Higher is better Pass³ · % · Higher is better Average assistant turns · turns · Lower is better