跳过主要内容
评测来源/SWE-Bench Verified
观察中编程

SWE-Bench Verified

SWE-bench · SWE-Bench Verified 的公开模型评测结果。

官方评测
来源说明

该来源测量模型在「编程」领域的具体交付表现，原始成绩会和统一能力读数分开展示。

限制: 不同版本、推理设置和测试集可能导致分数变化，请结合数据时间阅读。

使用方式

参与排名的来源会影响能力分；参考、交叉参考和观察来源仅用于对比与核验。

本来源的模型成绩

Claude Fable 59595.02026-08-05
Claude Opus 4.888.688.62026-08-05
Claude Sonnet 585.285.22026-06-30
Ornith 1.0 397B82.482.42026-08-05
DeepSeek V4 Pro80.680.62026-08-05
MiniMax-M380.580.52026-06-01
Qwen3.7 Max80.480.42026-05-19
Kimi K2.680.280.22026-08-05
MiniMax-M2.779.979.92026-06-01
DeepSeek V4 Flash7979.02026-08-05
MiMo-V2.5-Pro78.978.92026-08-05
Hy37878.02026-08-05
Mistral Medium 3.577.677.62026-08-05
Mistral Medium 3.577.677.62026-08-05
Qwen3.6 27B77.277.22026-08-05
Step 3.7 Flash76.576.52026-05-29
Qwen3.5 397B-A17B76.476.42026-08-05
MiniMax-M2.575.875.82026-08-05
Ornith 1.0 35B75.675.62026-08-05
Step 3.5 Flash74.474.42026-08-05
Hy3 preview74.474.42026-08-05
MiniMax-M2.17474.02026-08-05
GLM-4.773.873.82026-08-05
Qwen3.6 35B-A3B73.473.42026-08-05
GLM-572.872.82026-08-05
Qwen3.5 27B72.472.42026-08-05
Qwen3.5 122B-A10B7272.02026-08-05
MAI-Code-1-Flash71.671.62026-08-05
Laguna XS 2.170.970.92026-07-02
Kimi K2.570.870.82026-08-05
Nemotron 3 Ultra 550B A55B70.770.72026-06-04
MiniMax-M269.469.42026-08-05
Ornith 1.0 9B69.469.42026-08-05
North Mini Code67.667.62026-06-09
GLM-4.7-Flash59.259.22026-08-05

常见问题