THE CRUNCH
DeepSeek has open-sourced DeepInfer, the internal inference framework that serves its V4.1-Flash model, under the Apache License 2.0. Inference is the stage where a trained AI model handles requests and generates responses, and DeepInfer is the serving software that lets V4.1-Flash answer quickly at scale. The framework builds on vLLM with optimised compute kernels, distributed processing across multiple chips and faster text generation techniques.
The release also includes deepseek-recipe, which replicates the token processing of the official DeepSeek API so that self-hosted setups behave the same way. Code for future DeepSeek models will ship as open source too. The company said on X that the goal is "to help the community lower compute costs, reduce energy use, and make AI more affordable".
DeepSeek's models are already priced very low, and DeepInfer closes the remaining gap for running them even more cheaply. The Decoder suggests the move also adds pricing pressure on Western providers such as OpenAI and Anthropic, and could be politically motivated given the AI race between China and the US.