跳到正文
RadarAI· X:Arena (@arena)·· 3 小时前同新闻AI 评分78

月之暗面开源 Kimi K3.1 Pro 权重,登顶 Agent Arena 开源模型第一

AI 导读

月之暗面在 Hugging Face 开源 Kimi K3.1 Pro 模型权重,Arena 称其现为 Agent Arena 开源模型第一、总榜第 13。该结果基于 17.5K+ 真实长周期智能体会话,Kimi K3.1 Pro(Max)净提升 +5.1%,较 Kimi K3(Max)提升 4.2 个百分点,并在 Work、Code、Chat 三个类别中均为开源第一。

同一新闻,精选展示《Moonshot AI 在 Hugging Face 开放 2.8T 参数 Kimi K3.1 Pro 权重》

正文 · 原文

Arena.ai· @arena · X·2026-11-05 05:09· 50 分钟前AI 评分70

AI 导读

月之暗面(Kimi)在 Hugging Face 开源 Kimi K3.1 Pro 模型权重,据 Arena 称其现为 Agent Arena 开源模型第一。

正文 · 原文

Exciting news: @Kimi_Moonshot has released the open weights for Kimi K3.1 Pro, now #1 among open models in Agent Arena!

Across 17.5K+ real-world, long-horizon agentic sessions from our global community of users, Kimi K3.1 Pro (Max) delivers +5.1% net improvement, ranking #1 open and #13 overall.

It ranks #1 among open models in all three categories: Work, Code, and Chat.

This is a +4.2 percentage point improvement over Kimi K3 (Max), with +0.9% (#6 open). It also improved across all signals. Among open models, it ranks:
- #1 in Bash Reliability (successful shell commands) with +7.9%: (up from K3 at #13 with −4.5%).
- #1 in Praise vs Complaint (sentiment in user reactions) with +7.7%: (up from K3 at #6 with +1.3%).
- #1 in Steerability (responding to user corrections) with +6.5%: (up from K3 at #5 with +0.4%).
- #4 in Confirmed Success (explicit “yes, that worked” from users) with +3.9%: (from K3 at #5 with +3.8%).

Congrats to @Kimi_Moonshot on this contribution to the open ecosystem!

引用Kimi.ai@Kimi_Moonshot
Releasing the model weights of Kimi K3.1 Pro: https://huggingface.co/moonshotai/Kimi-K3.1-Pro We'll keep advancing open intelligence.
在 X 查看被引用的帖子

来源:Arena.ai · x.com

查看事件全部

来源:RadarAI · aihot.news