DeepSeek open-sources its inference engine DeepInfer under Apache 2.0

DeepSeek has open-sourced DeepInfer, its internal inference framework, under Apache 2.0. It is the serving software that lets the V4.1-Flash model answer quickly at scale, adding optimized compute kernels, cross-chip distributed processing, and faster text generation on top of vLLM. DeepSeek says code for future models will also be released as open source. Released alongside is deepseek-recipe, which replicates the token processing pipeline of DeepSeek's official API so self-hosted deployments behave identically to the official API. DeepSeek says the goal is to help the community cut compute costs and energy use and make AI more affordable. For teams running their own inference, the pair removes the work of matching the official API's preprocessing logic; for developers on DeepSeek's hosted API, deepseek-recipe offers a reference for reproducing official behavior locally.