批评 GPT-6 发布：多项基准倒退，Luna 掉分

Bee @thtbee_原推发布于 2026.09.23

在 X 上打开原推

This is such a underwhelming release by OpenAI.. They've gone backward in a lot of benchmarks The price decrease is the only thing holding them by a thin string!! In Artificial Analysis Coding Agent Index GPT-6 Sol (max) gains 2 points at half the Cost per Task of 5.6 and suprisingly😭GPT-6 Luna (max) drops by 2 points going behind Gemini 3.8 Flash..

这条是引用着另一条写的：x.com/thtbee_/status/2102467632985481225