国际权威大模型排行榜LMArena宣布，来自月之暗面的万亿参数开源模型Kimi K2接棒DeepSeek-R1-0528，成为全球排名第一的开源模型。K2 最大亮点是“智能体”（Agent）能力，可自动写代码、调 API、运行 Shell 命令、编辑 ...

China's free-for-all AI models, developed by firms like DeepSeek and Alibaba, present a viable alternative to US ...

OpenAI is implementing a major security overhaul with biometric access and offline systems, a response to allegations of IP theft and corporate espionage by Chinese rival DeepSeek.

对此，华为诺亚方舟实验室昨日发布声明回应，强调盘古 Pro MoE 模型是在昇腾硬件平台上独立开发和训练的基础大模型，研发过程未基于其他厂商的模型进行增量训练。 盘古 Pro MoE 开源模型是基于昇腾硬件平台开发、训练的基础大模型，并非基于其他厂商模型增量训练而来，在架构设计、技术特性等方面做了关键创新，是全球首个面向昇腾硬件平台设计的同规格混合专家模型，创新性地提出了分组混合专家模型（MoGE ...

German firm TNG has released DeepSeek-TNG R1T2 Chimera, an open-source variant twice as fast as its parent model thanks to a new 'Assembly-of-Experts' merge technique.

距离中国AI初创公司DeepSeek（香港高瓴资本管理公司旗下）发布其热门开源模型DeepSeek R1-0528的最新版本仅一个多月时间。

DeepSeek and other international rivals like ChatGPT have proven that being open source can lead to a more impressive LLMs, and still be lucrative.

DeepSeek R2 Launch Stalled as CEO Balks at Progress: Report R2, a successor to DeepSeek's wildly popular R1 reasoning model, was planned for release in May.

DeepSeek has delayed the launch of DeepSeek R2 following the new round of import bans impacting Nvidia chips.