DeepSeek Pledges to Publicly Release AI Model Code, Doubles Down on Open Source

Chinese startup DeepSeek announced today (February 21) that it will publicly release the code of its models, pledging to double down on open-source artificial intelligence.

Writing in a post on social media platform X, DeepSeek said it would open source five code repositories next week, describing the move as “small but heartfelt progress” that it would share with full transparency. “These modest building blocks of our online service have been documented, deployed, and tested in production,” the post said.

DeepSeek pledges to make AI model code public, doubling down on open source.

DeepSeek shook the global AI industry last month when it released its open-source R1 inference model, which rivals Western systems in performance while being developed at a much lower cost.

The company’s commitment to open source sets it apart from most AI companies in China, which, like their American rivals, lean toward closed-source models. DeepSeek founder Liang Wenfeng said in a rare interview with a Chinese media outlet last year that commercializing AI models is not a priority and that open source has the soft power to gain.

“Having others follow your innovation gives a great sense of accomplishment. In fact, open source is more cultural than commercial, and it helps us gain respect,” Liang said last July.

The newly released open source code will provide the infrastructure to support the AI models that DeepSeek shares publicly, building on top of those existing open source modeling frameworks.

The announcement comes after DeepSeek released a new algorithm called Native Sparse Attention (NSA) on Tuesday, designed to make long-tail contextual training and inference more efficient.

DeepSeek’s user base has exploded since last month. According to Aicpb.com, a Chinese website that tracks AI products, DeepSeek’s product is the most popular chatbot service in the Asian country with 22.2 million daily active users as of January 11, surpassing Douban’s 16.95 million users.

“At a meeting earlier this week, Chinese President Xi Jinping warmly welcomed DeepSeek founder Liang Wenfeng and gave the young expert a coveted seat alongside the leaders of the country’s largest private companies, showing Beijing’s willingness to support the company,” said Huiyao Wang, founder and president of the Center for China and Globalization.

“DeepSeek represents what Beijing wants, with “a new quality manufacturing force that will drive China forward,” Mr. Wang added, referring to a strategy launched by President Xi Jinping last year that bets on technological breakthroughs to drive growth and productivity gains across the economy .

Goldman Sachs estimates that China's GDP will increase by 20 to 30 basis points in the long term - by 2030 - and it expects the country's economy to start reflecting the positive impact of AI adoption as early as next year as AI-driven automation drives productivity improvements.

Source: https://www.baogiaothong.vn/deepseek-cam-ket-cong-khai-ma-mo-hinh-ai-tang-gap-doi-nguon-mo-192250221180736854.htm