Home Model series Claude Haiku Claude Haiku 4.5 Terminal-Bench Claude Haiku 4.5 · Terminal-Bench (版本未注明;Claude Haiku 4.5 发布材料所列) 41% · Main text evidence (ev-h45-terminal) Claude Haiku 4.5 · Terminal-Bench (版本未注明;Claude Haiku 4.5 发布材料所列) 41% · Main text evidence (ev-h45-terminal) Official source for Terminal-Bench · 版本未注明;Claude Haiku 4.5 发布材料所列 in Claude Haiku 4.5. 发布表报告值 · 41%. See “Introducing Claude Haiku 4.5” at Terminal-Bench 行;本地留存副本 line 50. Evidence ev-h45-terminal.
Source excerpt and location Claude Haiku 4.5 · Terminal-Bench (版本未注明;Claude Haiku 4.5 发布材料所列), configuration 发布表报告值, Score 41% (exact score publication date unconfirmed). See “Introducing Claude Haiku 4.5” at Terminal-Bench 行;本地留存副本 line 50. View score conditions
Terminal-Bench, version not shown — Haiku 4.5: 41.0%; Terminus 2; 11 runs (6 without thinking at 40.21%, 5 with 32K thinking at 41.75%). Official material Introducing Claude Haiku 4.5 · section located Location Main text · Terminal-Bench 行;本地留存副本 line 50 Material version 官方发布页;本地保存 HTML/TXT 与主表 PNG:本地留存副本、本地留存副本、本地留存副本;原文发布日期 2025-10-15,当前修订时间未知 Version published 2025-10-15 Disclosure date Unconfirmed · Launch or later disclosure unconfirmed Collected 2026-11-11 Historical basis No reliable historical version evidence available Citation text Claude Haiku 4.5 · Terminal-Bench (版本未注明;Claude Haiku 4.5 发布材料所列), configuration 发布表报告值, Score 41% (exact score publication date unconfirmed). Official source: Introducing Claude Haiku 4.5 (Main text, 官方发布页;本地保存 HTML/TXT 与主表 PNG:本地留存副本、本地留存副本、本地留存副本;原文发布日期 2025-10-15,当前修订时间未知), Terminal-Bench 行;本地留存副本 line 50, https://www.anthropic.com/news/claude-haiku-4-5#benchmarks. Benchlog https://benchlog.lekshome.cc/evidence/claude-haiku-4-5/ev-h45-terminal/?lang=en
Related scores and conditions Claude Haiku 4.5 · Terminal-Bench scores and configurations Metric and score Configuration and evaluation conditions Evidence relation Score 41% Reported 41.0% · Unit % Published Unconfirmed
发布表报告值 Settings SWE-bench: full 500 problems, 50 trials, no test-time compute, 128K thinking budget, default sampling; Terminal-Bench: Terminus 2, 11 runs split between no thinking and 32K thinking; OSWorld: OSWorld-Verified, 100 steps, 4 runs; GPQA/MMMLU/MMMU follow source table. Unknowns AIME 的 Python 工具使用情况按两个分列配置保存;其余各项的逐行配置以主文脚注为准。 Tools Tool access not recorded Reasoning budget not recorded Framework Terminus 2 Protocol Terminus 2; 11-run average, 6 no-thinking + 5 with 32K thinking。 Conditions 该来源未给 Terminal-Bench 版本;不映射到现有 2.0/2.1/4.0。 Directly supports this score
View this score in rankings