Home Model series Claude Haiku Claude Haiku 4.5 SWE-bench Verified Claude Haiku 4.5 · SWE-bench Verified (Verified subset (500 problems)) 73.3% · Main text evidence (ev-h45-swe) Claude Haiku 4.5 · SWE-bench Verified (Verified subset (500 problems)) 73.3% · Main text evidence (ev-h45-swe) Official source for SWE-bench Verified · Verified subset (500 problems) in Claude Haiku 4.5. 发布表报告值 · 73.3%. See “Introducing Claude Haiku 4.5” at SWE-bench Verified 行;本地留存副本 line 49. Evidence ev-h45-swe.
Source excerpt and location Claude Haiku 4.5 · SWE-bench Verified (Verified subset (500 problems)), configuration 发布表报告值, Pass rate 73.3% (exact score publication date unconfirmed). See “Introducing Claude Haiku 4.5” at SWE-bench Verified 行;本地留存副本 line 49. View score conditions
SWE-bench Verified — Haiku 4.5: 73.3%; full 500 problems, 50 trials, no test-time compute. Official material Introducing Claude Haiku 4.5 · section located Location Main text · SWE-bench Verified 行;本地留存副本 line 49 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 · SWE-bench Verified (Verified subset (500 problems)), configuration 发布表报告值, Pass rate 73.3% (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), SWE-bench Verified 行;本地留存副本 line 49, https://www.anthropic.com/news/claude-haiku-4-5#benchmarks. Benchlog https://benchlog.lekshome.cc/evidence/claude-haiku-4-5/ev-h45-swe/?lang=en
Related scores and conditions Claude Haiku 4.5 · SWE-bench Verified scores and configurations Metric and score Configuration and evaluation conditions Evidence relation Pass rate 73.3% Reported 73.3% · 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 Tools allowed Reasoning budget Thinking budget 128K Framework not recorded Protocol SWE-bench Verified full 500-problem set; 50 trials; no test-time compute。 Conditions 128K thinking budget; default sampling。 Directly supports this score
View this score in rankings