PANews reported on January 31 that according to Jinshi, NVIDIA (NVDA.O) website shows that DeepSeek R1 671b has been released as a preview version of NVIDIA NIM microservices on build.nvidia.com. DeepSeek R1 NIM microservices can process up to 3,872 tokens per second on a single NVIDIA HGX H200 system. Developers can test and experiment with the API, which is expected to be launched soon as part of NVIDIA's AI Enterprise software platform in the form of a downloadable NIM microservice.