In a statement posted on X on Friday, the Hangzhou-based company revealed that starting next week, its "tiny team" dedicated to exploring artificial general intelligence will begin "open-sourcing five repositories, sharing our modest yet genuine progress with complete transparency."
A code repository serves as a storage space where developers can access and contribute to software projects. Open-source repositories are often hosted on well-known centralized platforms such as Microsoft’s GitHub.
While code repositories can house significant company assets, the open-source approach of DeepSeek’s V3 and R1 models, which allows free usage and modification, has been pivotal to the start-up's rapid ascent in recent months.
“As part of the open-source community, we believe that every line shared becomes collective momentum that accelerates the journey,” the company wrote. “Daily unlocks are coming soon,” it added. “No ivory towers – just pure garage-energy and community-driven innovation.”
The company has been making waves globally after its impressive, budget-friendly large language models (LLMs) surprised the tech world. Following the buzz around its models that caused AI stocks to tumble last month, DeepSeek has been flying under the radar, mostly sticking to technical research without much public activity.
This week, DeepSeek released a study featuring founder and CEO Liang Wenfeng as one of the 15 co-authors. The paper focused on “native sparse attention,” aimed at boosting LLM efficiency when handling large data sets. This announcement highlighted DeepSeek’s dedication to open research and development.
While open-source code can come with various licensing options, it’s often available for free for personal and commercial use.
DeepSeek’s advancements in open-source models have ignited discussions in the AI community, especially since leading developers like OpenAI have kept their top technologies under wraps.
The announcement from DeepSeek on Friday received thousands of likes on X, with a Reddit user in a forum for local AI models calling the company a “gift to humanity.”
Earlier this week, the start-up also updated its business registry to reflect a broader operational scope, which analysts believe could lead to monetizing its LLMs.