各榜单最高分GLM 5.3 (v3)Step 5 Preview
开发方ZhipuStepFun
Artificial Analysis Intelligence Index v4.344.843.7
Terminal-Bench 4.041.8—
Terminal-Bench 4.0 (AA)41.933.3
Code Arena | WebDev Overall1622.01570.0
Agent Arena · Overall (Ranking / Pareto)2.6—
Artificial Analysis Coding Agent Index v1.553.6—
DeepSWE v1.169.0—
雷达 IQ(Codex)96.7—
最低真实单价 / 百万 token$0.0089$0.0036

数据:real-api-pricing数据:codexradar来源与许可