DeepSeek announced it will further open-source its internal inference framework, DeepInfer. The company stated that it has achieved high-throughput, low-latency large-scale inference services for DeepSeek V4.1 Flash based on the framework. Going forward, DeepSeek confirmed that inference code for all new models and subsequent optimizations will be continuously updated within the DeepInfer repository. This move extends DeepSeek's strategy of making its core technical infrastructure publicly accessible to developers and researchers.