Lenovo released the DeepSeek all-in-one solution

According to Zhitong Finance APP, on February 5th, Lenovo Group (00992) and Muxi Co., Ltd. jointly released an all-in-one solution based on the DeepSeek large model. This solution is centered around the architecture of "Lenovo server/workstation + Muxi training and 推 integrated GPU + self-developed algorithms," combined with the Lenovo AI Force intelligent agent development platform, launching dual product forms of intelligent agent all-in-one machines and training 推 integrated servers. It is the first to achieve full-chain coverage from training large models with hundreds of billions of parameters to scenario-based inference implementation, providing new pathways for enterprises to solve three core pain points: complex computing power deployment, high technical thresholds, and difficulties in security and control.

In recent days, DeepSeek has attracted widespread global attention due to its low-cost, high-performance large model research and development achievements. The company's latest R1 model has already matched the capabilities of top industry models.

Against this backdrop, Lenovo has innovatively launched two major products: the DeepSeek intelligent agent all-in-one machine for agile deployment, which uses the ThinkStation PX workstation as its carrier and is equipped with the Muxi Xisi N260 GPU. Test data shows that under the same concurrent conditions, the Muxi Xisi N260 GPU achieves 110%-130% of the inference performance of the NVIDIA L20 GPU in the Qwen2.5-14B model, supporting local deployment of various parameter distilled models for DeepSeek. The DeepSeek training 推 integrated machine, aimed at broader scenarios, is based on the Lenovo Wenti WA5480 G3 AI server and is equipped with 8 Xiyun C500 professional GPUs, providing strong computing power support for large model training and inference.

Intelligent agent all-in-one machine based on Lenovo ThinkStation PX

Training 推 integrated machine based on Lenovo Wenti WA5480 G3 AI server

The breakthrough in the software ecosystem is also noteworthy. The Lenovo AI Force intelligent agent development platform reconstructs the enterprise-level AI development paradigm through three technological innovations: "integrated integration, private deployment, and easy operation and maintenance." In terms of "integrated integration," the platform deeply integrates mainstream large model libraries such as DeepSeek, providing pre-set industry knowledge graphs and standardized API interfaces. Through a visual drag-and-drop interface, enterprises can shorten the business system docking cycle from weeks to hours; the "private deployment" solution ensures data security and compliance, with high customization and redevelopment capabilities, supporting seamless integration with existing ERP and MES systems while retaining customized development interfaces The "easy operation and maintenance" feature allows users to use it out of the box, significantly reducing hardware procurement and operation and maintenance costs for enterprises.

Lenovo AI Force Intelligent Agent Development Platform

Gu Xu Guang, Senior Director of Lenovo China's Solution Service Business Group "Qing Tian" AI Technology Center, emphasized, "Lenovo provides an intelligent integrated software and hardware platform, with the Muxi GPU providing the hardware foundation, and the DeepSeek model constructing the intelligent core, achieving trusted hardware across the entire chain. In actual programming development scenarios, the code generation function based on DeepSeek achieves an adoption rate of 35% or more."

Lü Jian Fei, General Manager of Muxi, added, "Muxi and Lenovo will continue to invest resources and R&D efforts to continuously optimize the inference efficiency and training performance of large models, and provide customized application templates for more vertical industry scenarios. Both parties also plan to collaborate with industry partners and the developer community to promote broader and deeper implementations."