Recently, DeepSeek, a leading Chinese AI large model, was deployed and launched at the Dongguan Artificial Intelligence Large Model Center, which is set to commence operations soon. As China's first city-level large model center dedicated to the manufacturing sector, and built on Huawei's software and hardware foundation, the Center has established 100P of domestically produced intelligent computing power, three foundational large models, and two industry-specific models for manufacturing. It offers AI intelligent computing services and application models to enterprises citywide, empowering traditional industries and manufacturing businesses to transform and upgrade, thereby boosting production efficiency.

Dongguan is keeping pace with the AI development trend by deploying the full DeepSeek-R1 671B model at the Center. Users can log in and immediately access the full-version services. This deployment not only supports businesses in building NLP-based application scenarios but also provides users with powerful capabilities for developing AI applications.

The domestically-produced AI computing power foundation built at the Center provides high-performance computing support for the Pangu Large Models and DeepSeek, enabling efficient training and inference. Pangu Large Models, with its industry-specific expertise, supports the scenario applications of DeepSeek, while DeepSeek leverages advanced distillation technology and efficient inference capabilities to help improve the performance of the Pangu Large Models. The complementary advantages of both models optimize training efficiency, achieve knowledge transfer, and bring breakthroughs to the industry.

In the future, the Center will deepen its collaboration with companies like Huawei, utilizing techniques such as model distillation and fine-tuning. By integrating the industry experience of Pangu Large Models with the efficient and cost-effective advantages of DeepSeek, the Center aims to implement more intelligent scenarios in the manufacturing sector. Key areas such as intelligent quality inspection, supply chain optimization, and process design will benefit, allowing AI technology to further empower Dongguan's manufacturing industry.

Source: Lingnan on the Cloud

DeepSeek正式登陆东莞

近日，国内领先的人工智能大模型DeepSeek在东莞市人工智能大模型中心部署上线，大模型中心也将于近期开服。东莞市人工智能大模型中心是全国首个面向制造领域的城市级大模型中心，依托华为软硬件基础底座，已建成100P国产化智能算力、3个基础大模型、2个制造行业大模型。它能为全市企业提供AI智能算力服务和AI应用场景模型，助力传统产业与制造企业转型升级，提高生产效率。

此次东莞紧跟AI发展潮流，在大模型中心完成DeepSeek-R1 671B满血模型部署，用户登录即可享受满血版服务。这不仅支持企业构建基于NLP的场景应用，还赋予用户强大的人工智能应用开发能力。

中心建成的国产化人工智能算力底座，为盘古大模型和DeepSeek提供高性能算力，支撑高效训推。盘古大模型凭借行业积累，为DeepSeek的场景应用提供支撑；DeepSeek则利用先进蒸馏技术和高效推理能力，助力盘古大模型提升性能。二者优势互补，优化了训练效率，实现知识迁移，为行业带来新突破。

后续，东莞市人工智能大模型中心还将与华为等企业深化合作，运用模型蒸馏、微调等技术，融合盘古大模型的行业经验和DeepSeek高效低成本的优势，推动制造领域智能场景的落地，如智能质检、供应链优化、工艺设计等关键领域，让AI技术更深入地赋能东莞制造业。

文 | 记者 余晓玲 通讯员 莞宣
译 | 林佳岱
审 | 王枥焓