各榜单最高分Claude Sonnet 5.5GPT-6 Astra
开发方AnthropicOpenAI
Artificial Analysis Intelligence Index v4.356.052.7
Terminal-Bench 4.0—58.2
Terminal-Bench 4.0 (AA)63.659.6
Code Arena | WebDev Overall1709.01789.0
Agent Arena · Overall (Ranking / Pareto)—12.2
Artificial Analysis Coding Agent Index v1.568.461.6
OpenDesign Arena · Quality ranking86.882.7
DeepSWE v1.1—74.1
雷达 IQ(Codex)—110.4
最低真实单价 / 百万 token$0.0027$0.052

数据:real-api-pricing数据:codexradar来源与许可