跳过主要内容
评测来源/ProgramBench
参与排名编程

ProgramBench

Vals AI · ProgramBench 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 Opus 5.518.5100.02026-11-05
Claude Fable 5.1779.52026-11-05
Claude Sonnet 5.56.576.92026-11-05
GPT-6 Astra5.571.72026-11-05
Claude Opus 5358.72026-11-05
GPT-6.1 Sol358.72026-11-05
Muse Spark 1.32.556.12026-11-05
Gemini 4 Argon2.556.12026-11-05
Claude Haiku 5.52.556.12026-11-05
Claude Fable 5253.52026-11-05
Kimi K3253.52026-11-05
GPT-6 Sol253.52026-11-05
GPT-5.6 Sol1.550.92026-11-05
GLM 5.31.550.92026-11-05
Claude Opus 4.8148.32026-11-05
Grok 4.6148.32026-11-05
Gemini 3.8 Flash148.32026-11-05
MiMo V2.6 Flash0.545.72026-11-05
MiMo V2.6 Flash0.545.72026-10-18
MiMo V2.6 Flash0.545.72026-10-17
Claude Sonnet 4.60.545.72026-11-05
GPT-5.40.545.72026-11-05
GPT-5.50.545.72026-11-05
GLM-5.20.545.72026-11-05
GPT-5.6 Terra0.545.72026-11-05
Inkling Small0.545.72026-11-05
Muse Spark 1.20.545.72026-11-05
DeepSeek V4.1 Flash0.545.72026-11-05
Grok 4.70.545.72026-11-05
MiMo V2.6 Pro0.545.72026-11-05
GPT-6 Luna0.545.72026-11-05
Claude Haiku 4.5043.12026-11-05
Gemini 3.1 Flash Lite Preview043.12026-11-05
GPT-5.4 mini043.12026-11-05
MiniMax-M2.7043.12026-11-05
Qwen3.6 Plus043.12026-11-05
GLM-5.1043.12026-11-05
Muse Spark 1.1043.12026-11-05
Claude Opus 4.7043.12026-11-05
Grok 4.3043.12026-11-05
Kimi K2.6043.12026-11-05
Laguna M.1043.12026-11-05
Laguna XS.2043.12026-11-05
Kimi K2.7 Code043.12026-11-05
Claude Sonnet 5043.12026-11-05
Grok 4.5043.12026-11-05
GPT-5.6 Luna043.12026-11-05
Inkling043.12026-11-05
Gemini 3.5 Flash Lite043.12026-11-05
Gemini 3.6 Flash043.12026-11-05
DeepSeek V4 Flash 0731043.12026-11-05
Qwen3.8 Max043.12026-11-05
DeepSeek V4 Pro 0813043.12026-11-05
GLM-5.3 Flash043.12026-11-05
GLM-5.3 Flash043.12026-10-17
GLM-5.3 Flash043.12026-10-18

常见问题