Home Model series Claude Opus Claude Opus 4.6 Claude Opus 4.6 · Terminal-Bench · 2.0;代理框架按协议区分 · Scores and configurations Claude Opus 4.6 · Terminal-Bench · 2.0;代理框架按协议区分 · Scores and configurations Claude Opus 4.6 record for Terminal-Bench · 2.0;代理框架按协议区分. 3 score records for this release, including 3 exact values (may include historical records). Includes 2 source locations.
Scores and evaluation conditions Claude Opus 4.6 · Terminal-Bench scores and configurations Metric and score Configuration and evaluation conditions Sources Pass rate 65.4% Reported 65.4% · Unit % Published Unconfirmed
Terminal-Bench 2.0 · Max Settings Harbor scaffold、Terminus-2 harness、default parser;adaptive thinking;Max effort。 Unknowns none recorded Tools Tool access not recorded Reasoning budget max Framework Harbor scaffold; Terminus-2 harness; default parser Protocol Terminal-Bench 2.0;Harbor scaffold;Terminus-2 harness;default parser。 Conditions 89 tasks;15 runs/task;1,335 trials;3 个不同时间批次;adaptive thinking、max effort。节点 n2-standard-32,1×资源配额、3×硬抢占上限、1× timeout;结果不与其他 harness 的 GPT-5.2 Codex CLI 分值混同。 View this score in rankings Pass rate 55.1% Reported 55.1% · Unit % Published Unconfirmed
Terminal-Bench 2.0 · Low Settings Harbor scaffold、Terminus-2 harness、default parser;adaptive thinking;Low effort。 Unknowns 该 effort 点未单列试验次数。 Tools Tool access not recorded Reasoning budget low Framework Harbor scaffold; Terminus-2 harness Protocol Terminal-Bench 2.0;Harbor/Terminus-2。 Conditions Figure 2.5.A;adaptive thinking、Low effort;卡片未单列该点试验次数。 View this score in rankings Pass rate 61.1% Reported 61.1% · Unit % Published Unconfirmed
Terminal-Bench 2.0 · Medium Settings Harbor scaffold、Terminus-2 harness、default parser;adaptive thinking;Medium effort。 Unknowns 该 effort 点未单列试验次数。 Tools Tool access not recorded Reasoning budget medium Framework Harbor scaffold; Terminus-2 harness Protocol Terminal-Bench 2.0;Harbor/Terminus-2。 Conditions Figure 2.5.A;adaptive thinking、Medium effort;卡片未单列该点试验次数。 View this score in rankings
Benchmark versions and metrics Aliases Terminal-Bench 2.0
Pass rate · % · Higher is better Mean reward · % · Higher is better