GPT-6 2.2 Sol 与 Astra 发布,Astra 在 ARC-3 上满分
OpenAI 这波更新有点平淡:Sol 没什么进步,Astra 虽然把 ARC-3 基准刷满了,但那基准本来就快饱和了。而且俩模型都没赢过 Claude。想选模型的人可以看看实测对比再决定。
OpenAI 发布 GPT-6 2.2 系列的两个模型:Sol 和 Astra。Astra 在 AGI ARC-3 基准上取得 100% 满分成绩,不过该基准此前已接近饱和,含金量存疑。Sol 则被认为相比前代没有明显升级。
值得注意的是,这两个模型在对比评测中都没有跑赢对应的 Claude 模型。这一结果延续了近期 Claude 系列在多类基准上的领先态势,也说明单看某个饱和基准的分数已难以区分模型实力。
GPT-6 2.2 Sol and Astra released. Sol doesn't seem to be a huge upgrade, but Astra has 100%ed the AGI ARC-3 benchmark, which was already near saturation anyway. Neither have beaten their Claude companions, though.
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, OpenAI, Claude, ARC-3.
Source attribution: curated from @koltregaskes (@koltregaskes).