DeepSeek Will Open Source Five Code Libraries

On February 21, DeepSeek announced that it will gradually open source five code libraries starting next week, unlocking new content each day for the next five days.

In the post, DeepSeek positioned itself as a “small team” dedicated to exploring artificial general intelligence (AGI). Regarding the upcoming open-source code libraries, it will share new progress without reservation. The company stated that the current online service module has completed comprehensive testing, deployment has been perfected, and it has the capability to be put into production environments. Although the specific contents of the code library have not been detailed, based on DeepSeek’s past technology releases, these open-source contents may cover machine learning models, data processing pipelines, and related frameworks, providing effective support for developers in various projects. DeepSeek also emphasized, “As a member of the open-source community, we believe that every line of shared code will converge into a force driving progress.” It claimed that in this field, there are no high barriers, only pure entrepreneurial passion and community-driven innovation.

The company’s open-source initiative has had a profound impact on the entire AI industry, not only lowering technical barriers, allowing small and medium-sized enterprises and developers to more easily use large model technology, develop innovative applications, but also encouraging community sharing, enhancing cooperation and communication among users, and promoting the rapid evolution of AI technology.

The latest data from the data analysis platform QuestMobile shows that from its launch to February 9, the DeepSeek App has been downloaded more than 110 million times, with a peak weekly active user scale of nearly 97 million. Among them, from January 20 to January 26, the weekly download volume of the DeepSeek App reached 2.26 million times. The following week, the download volume soared to 63 million times, an increase of over 2700% week-on-week. Under the influence of DeepSeek, some large model companies that have previously firmly followed the open-source route are gradually changing their attitudes.

On February 17, Sam Altman, CEO of OpenAI, posted on X, asking everyone what they would like the next open-source project to be: a “fairly small but still needs to run on a GPU o3-mini level model,” or “the best possible phone-sized model.” This is seen by the outside world as a signal that a large model will soon be open-sourced, and it would be the company’s second open-sourcing of a large model after GPT-2.

SEE ALSO: Baidu: ERNIE’s 4.5 Series Will Be Open Source From June 30th