Huawei and SiliconFlow Collaborate to Make DeepSeek AI Models Accessible and Affordable

Amidst the Lunar New Year holidays, Huawei Technologies and its cloud-computing division worked diligently with local Beijing-based AI infrastructure startup SiliconFlow. Their collaboration aimed to provide end users efficient and affordable access to DeepSeek’s advanced artificial intelligence (AI) models through Huawei’s platform.

The Partnership

Huawei’s cloud unit partnered with SiliconFlow to offer DeepSeek’s large language model V3 and reasoning model R1 to end users via Huawei’s Ascend cloud service. According to Huawei Cloud, the performance of these models on Ascend matches their performance on global premium graphic processing units (GPUs).

Affordable Pricing

According to SiliconFlow, the start-up responsible for hosting the DeepSeek models, accessing the large language model V3 on its platform now costs 1 yuan (US$0.13) for 1 million input tokens and 2 yuan (US$0.27) for 1 million output tokens. The charges for the reasoning model R1 are 4 yuan (US$0.53) and 16 yuan (US$2.12) per 1 million input and output tokens respectively. Such pricing ensures affordability while maintaining top-tier performance.

Global Competition in AI Infrastructure

The rise in popularity and demand for open-source DeepSeek models in both China and the US underscores a growing trend in the industry. Microsoft, an investor in OpenAI, recently introduced support for DeepSeek’s R1 model on Azure, its cloud-computing platform as well as on GitHub. This development allows developers to build applications leveraging R1’s capabilities directly on Microsoft’s Copilot+ personal computers.

Similarly, Amazon Web Services has enabled developers to harness the potent yet cost-effective R1 to facilitate application creation. These announcements highlight the intense competition in providing powerful, accessible AI infrastructure services to meet the demands of both businesses and consumers.

Huawei’s Ascend Cloud Service

Huawei’s Ascend cloud service utilizes proprietary Ascend solutions for compute resources, potentially featuring a diverse range of hardware elements such as Huawei’s own server clusters, AI modules, and accelerator cards. These devices and technological solutions aim to deliver reliable and high-performance cloud services tailored to AI applications.

Despite the details of the specific chips used in Ascend cloud service remaining undisclosed by Huawei and SiliconFlow, the collaboration leverages Huawei’s advancements in AI hardware to offer users exceptional performance at competitive prices.

The Future of AI in Cloud Services

The collaboration between Huawei and SiliconFlow represents a significant step forward in democratizing access to cutting-edge AI technology. By offering DeepSeek’s models through its Ascend cloud service at a highly affordable price, Huawei is positioning itself as a leading player in the competitive AI infrastructure landscape.

Furthermore, this strategic partnership showcases the potential for international collaboration in the realm of AI, where companies spanning different regions can leverage each other’s strengths to deliver superior products and solutions to the global market.

We invite readers to share their thoughts on this exciting development in the cloud computing and AI industry. Feel free to comment below or subscribe to our platform to stay informed about the latest updates on technology and innovation.