Chinese artificial intelligence startup DeepSeek has officially released its internally developed inference framework, DeepInfer, as an open-source tool. The framework is specifically designed to optimize the performance of the company’s DeepSeek-V4.1-Flash model. According to the company, the tool is compatible with both Nvidia’s SM100 Blackwell graphics processing units (GPUs) and hardware from Huawei, signaling an effort to maintain operational flexibility amidst ongoing international chip export restrictions.

The release is part of a broader company strategy to promote open-source technology. A senior researcher at DeepSeek characterized the current industry landscape as a ideological struggle between open and closed-source development, describing the divide as a battle between “light and darkness.” By making the framework public, DeepSeek aims to lower the barriers for developers to run high-performance AI models, potentially challenging the market share currently held by companies that keep their proprietary software locked behind closed systems. While the company frames this as a pursuit of "total victory" for open-source AI, the move also serves as a practical solution to hardware compatibility challenges in the Chinese market.