Artificial Analysis 指出，Inception 发布 Mercury 3 时提到的 AA-Omniscience 和 GDPval-AA v2 分数并非其 Intelligence Index，也不遵循其方法论，无法与 Artificial Analysis 上的分数比较。AA-Omniscience 使用私有数据集，仅开源了 10% 的公开子集；GDPval-AA v2 需采用其评分方法和模型对比集才具可比性。被引用的 Inception 公告称 Mercury 3 是迄今最强的扩散 LLM，在多项基准上超过 GPT-6 Luna 和 Claude Haiku 5.5，速度超过 1,000 tokens/s，规模为 Mercury 1 的 40 倍以上，上下文窗口 1M token，已在 Inception API、Baseten 和 OpenRouter 上线。
Note that this is not our Intelligence Index. The announcement also referred to scores for AA-Omniscience and GDPval-AA v2, which do not follow our methodology and are not comparable to scores on Artificial Analysis.
AA-Omniscience uses a private dataset (though we open sourced a public subset of 10%) and GDPval-AA v2 needs to use our grading methodology and our model comparison set to be comparable.
Mercury 3 is live: the most capable diffusion LLM released to date. It outperforms GPT-6 Luna and Claude Haiku 5.5 across a range of benchmarks while running at over 1,000 tokens per second. More than 40x the size of Mercury 1, 3x more intelligent across standard benchmarks, and just as fast, with a 1M-token context window. Available today on the Inception API, @baseten, and @OpenRouter .在 X 查看被引用的帖子
来源：Artificial Analysis · x.com