AI EVALUATION, IN CONTEXT.BY EQUALITY MACHINE

Model reports

SWE-bench Pro

Harness and split version must be recorded with any score.

评测信息与证据

来源: Model reports
发布者: 未收录
评测发布日期: 2025-09-23
采集日期: 未收录
模态: 未收录

本站保留来源报告的结果，不代表独立复测。请先核对任务版本、评测协议、分数单位和报告日期；不同运行条件不能直接合并为统一排名。

原始资料

报告成绩 10

按来源中的观察顺序展示，顺序不代表排名。日期保留原始精度。

模型	数值	单位	协议	报告日期与口径	来源
Claude Opus 5	79.2	percent	adaptive thinking, max effort, averaged over 5 trials	2026-07-24 · document_publication	Claude Opus 5
DeepSeek-V4-Pro	52.1	percent	non-think, resolved	2026-04-22 · document_publication	DeepSeek-V4-Pro
DeepSeek-V4-Pro	54.4	percent	think high, resolved	2026-04-22 · document_publication	DeepSeek-V4-Pro
DeepSeek-V4-Pro	55.4	percent	think max, resolved	2026-04-22 · document_publication	DeepSeek-V4-Pro
DeepSeek-V4-Flash	52.6	percent	think max, resolved	2026-06-25 · document_publication	DeepSeek-V4
Gemini 3.1 Pro	54.2	percent	Thinking (High), Single attempt	2026-02-19 · document_publication	Gemini 3.1 Pro
Hy4 preview	65.7	percent	public split	2026-08-28 · document_publication	Hy4 preview
Grok 4.5	64.7	percent	resolve rate	2026-07-08 · document_publication	Grok 4.5
GLM-5.1	58.4	percent	as reported	2026-04-03 · document_publication	GLM-5.1
GLM-5.2	62.1	percent	OpenHands, tailored instruction prompt, temp 1, top_p 1, max_new_tokens 32k, 400K context	2026-06-16 · document_publication	GLM-5.2