Home Model series Claude Sonnet Claude Sonnet 5 Claude Sonnet 5 · Terminal-Bench · 2.1 · Scores and configurations Claude Sonnet 5 · Terminal-Bench · 2.1 · Scores and configurations Claude Sonnet 5 record for Terminal-Bench · 2.1. 1 score record for this release, including 1 exact value (may include historical records). Includes 2 source locations.
Scores and evaluation conditions Claude Sonnet 5 · Terminal-Bench scores and configurations Metric and score Configuration and evaluation conditions Sources Mean reward 80.4% Reported 80.4% · Unit % Published Unconfirmed
Xhigh effort Settings Xhigh effort in Terminal-Bench 2.1. Unknowns No other model sampling details are stated for that evaluation. Tools Tool access not recorded Reasoning budget xhigh Framework mini-SWE-agent harness Protocol Terminal-Bench 2.1; mini-SWE-agent harness; 89 tasks × 5 attempts = 445 trials. Conditions Xhigh effort; GKE with 1x timeout rate and 3x memory ceiling before pod preemption. View this score in rankings
Benchmark versions and metrics Aliases Terminal-Bench 2.1
Mean reward · % · Higher is better Score · % · Higher is better