DeepInfer DeepSeek's production inference framework, now fully open-sourced under Apache 2.0. It powers DeepSeek-V4.1-Flash at scale with high-performance kernels, distributed inference, speculative decoding (DSpark), structured outputs, and support for NVIDIA SM100 and Huawei Ascend 950. Also includes the companion deepseek-recipe for token encoding aligned with the official API.
Visit productFeatured media
Loading media…
View on XCosigns Received
“This is the kind of open source the AI ecosystem needs. Real production infrastructure, battle-tested at scale, not just another model release.”
“Amazing. For those who have used DeepSeek before, you know that the inference that they have on the official provider is absurdly good+cheap. This is a big release.”