Models

GPT-6.2 Astra 写作任务胜率从 91.2% 降至 85.9%

Why we picked it

看数据说话:Agent Arena 上用户实测,GPT-6.2 Astra 写作成功率比 GPT-6 Astra 掉了 5 个百分点以上,创意文案掉得最狠。日常靠它写文案的可以暂缓升级,先留在旧版观察。

Agent Arena 平台数据显示,OpenAI 的 GPT-6.2 Astra 在写作与翻译任务上的用户确认成功率为 85.9%,低于上一代 GPT-6 Astra 的 91.2%。

Body

下降主要集中在创意文案类请求,文档与研究类任务也有小幅下滑,但幅度较小。

这是用户投票确认的结果,反映新版本在生成类任务上的实际体验出现回退,对以文案产出为主要场景的用户有直接参考意义。

Original · lmarena.ai

Writing is where GPT-6.2 Astra by @OpenAI delivers worse results than the previous generation. On Agent Arena, users confirmed 85.9% of GPT-6.2 Astra’s writing and translation tasks as successful. For GPT-6 Astra it was 91.2%. The drop is sharpest in creative copy requests. Documents and research tasks also edge down, though by less.📊 1 ⚡

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include GPT-6.2 Astra, OpenAI, Agent Arena, writing.

Source attribution: curated from lmarena.ai (@lmarena_ai).