DeepSeek, a Chinese AI startup, will release some of its technology to the public next week.

Developers and researchers will gain access to five code repositories from its R1 platform, including the underlying code, data, and development methodologies.

Founded 20 months ago in Hangzhou, DeepSeek has drawn attention for AI models that recently outperformed OpenAI and Meta in benchmark tests.

This open-source initiative sets DeepSeek apart from companies like OpenAI, which have moved away from open releases.

While Meta has shared some models, DeepSeek’s move aims to promote wider adoption but has raised security concerns.

Founder Liang Wenfeng, who previously worked in a hedge fund, has not disclosed outside funding.

The company stressed that the shared resources are tested, deployed, and well-documented, reinforcing its commitment to transparency in AI research.

Recent DeepSeek developments