Kimi K3.1 Pro Alignment Index 并列第二,Moonshot 成排名最高的中国实验室
Moonshot 的 Kimi K3.1 Pro 在 LMArena 的 Alignment Index 拿到 78.2 分,和 Anthropic 并列第二,把阿里甩开 12 分。做意图遵循相关应用的开发者可以看看这份榜单,选模型时多一个参考维度。
LMArena 的 Alignment Index 排行榜更新,Moonshot AI 的 Kimi K3.1 Pro 拿到 78.2 分,与 Anthropic 并列实验室总分第二,置信区间与 Fable-5.2 重叠。
这一成绩使 Moonshot 成为该榜单上排名最高的中国实验室,比排在第七位的阿里巴巴高出 12 分。榜单同时显示,开源权重模型整体仍落后于闭源模型:排名最高的开源模型 Mimo-V2.6-Flash 位列第 26 名,其后依次是 Hy4、Beam、Inkling、DeepSeek 和 GLM。
Alignment Index 衡量模型遵循意图与指令一致性的表现,此前该榜单长期由美国闭源实验室主导,Kimi K3.1 Pro 的分数意味着中国实验室在一致性维度上首次进入头部梯队。
Alignment Index Top-10 Lab Kimi K3.1 Pro's alignment score puts @Kimi_Moonshot tied for #2 across labs. Moonshot and @Anthropic both score 78.2 with the confidence intervals overlapping with Fable-5.2. Kimi K3.1 Pro also makes Moonshot the highest-ranked Chinese lab on the index, 12 points ahead of Alibaba at #7 . The rest of open weight models are trailing closed releases. Mimo-V2.6-Flash, is #26 on the Index, followed by Hy4, Beam, Inkling, DeepSeek and GLM.📊 9 ⚡
This report is categorized under AITOP's "Models" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Kimi K3.1 Pro, Moonshot, Anthropic, Alignment Index.
Source attribution: curated from lmarena.ai (@lmarena_ai).