模型22

GPT-6.2 Astra 与 Sol 高推理档位表现更优

原标题:Both models rank higher on the Evidence-Based and Disclaims Assumptions traits at higher effort variants.

Artificial Analysis@ArtificialAnlys发布时间 (北京时间)

两款模型在更高推理档位下,在“基于证据”和“不预设假设”两项特质上的排名都更高。 ➤ 基于证据,从低到最高:GPT-6.2 Astra 12 至 54,GPT-6.2 Sol 16 至 68 ➤ 不预设假设,从低到最高:GPT-6.2 Astra 20 至 52,GPT-6.2 Sol 26 至 77

中文译文尚未完成,先为你展示原文。

Both models rank higher on the Evidence-Based and Disclaims Assumptions traits at higher effort variants.

➤ Evidence-Based, low to max: GPT-6.2 Astra 12 to 54, GPT-6.2 Sol 16 to 68
➤ Disclaims Assumptions, low to max: GPT-6.2 Astra 20 to 52, GPT-6.2 Sol 26 to 77

OpenAI / ChatGPT评测基准#评测/基准#OpenAI