PulseAugur
EN
LIVE 20:00:56
中文(ZH) 阿里云真武超节点适配Kimi K3

Alibaba Cloud and Kimi K3 Partner for Domestic Large Model Computing Power

Alibaba Cloud's Zhenwu M890 super node instances have been adapted to run the Kimi K3 large language model, marking the first time a nearly 3 trillion parameter model has been successfully operated on such a node in China. This collaboration aims to advance domestic computing power capabilities, with both parties planning to offer Kimi K3 model APIs through their respective platforms. The Zhenwu M890 utilizes high-speed interconnects to overcome the limitations of traditional AI clusters, enabling efficient operation of massive parameter models and supporting long context windows for complex tasks. AI

IMPACT This collaboration enhances domestic capabilities for running trillion-parameter models, potentially accelerating the development and deployment of large AI models in China.

RANK_REASON Partnership between a major cloud provider and a large language model to enable domestic large-scale AI model operation.

Read on 36氪 (36Kr) →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

Alibaba Cloud and Kimi K3 Partner for Domestic Large Model Computing Power

COVERAGE [2]

  1. 雷峰网 (Leiphone) TIER_1 中文(ZH) ·

    Alibaba Cloud Zhenwu Supernode Day 0 Adapts Kimi K3 Large Model, Both Parties Launch Domestic Computing Power Cooperation

    <p>7月28日,阿里云真武M890超节点实例已Day0适配Kimi K3,这是国内首个跑通近3万亿参数模型的超节点。双方将进一步展开国产算力合作,千问AI平台和阿里云百炼也将提供Kimi K3 的模型API。</p><p>&nbsp;</p><p>Kimi&nbsp;K3是目前业界参数规模最大的模型之一,跑通这一规模的模型需要将参数分散到几十张甚至上百张高速互联的GPU卡,让万亿参数“跑得又稳又快”。然而,普通AI集群因通信带宽的局限,无法满足高吞吐、低延迟、高并发的需求。</p><p>&nbsp;</p><p style="text-align: c…

  2. 36氪 (36Kr) TIER_1 中文(ZH) ·

    Alibaba Cloud Zhenwu Supernode Adapts Kimi K3

    36氪获悉,据阿里云消息,阿里云真武M890超节点实例已Day0适配Kimi K3,这是国内首个跑通近3万亿参数模型的超节点。双方将进一步展开国产算力合作,千问AI平台和阿里云百炼也将提供Kimi K3的模型API。