NVIDIA has announced the availability of new NVIDIA Inference Microservices (NIMs), including Mistral-7B, Mixtral-8x7B, and Mixtral-8x22B. These microservices are designed to provide optimized AI inference for large language models (LLMs). The new NIMs can be downloaded from the NVIDIA API Catalog. NVIDIA NIMs are part of NVIDIA AI Enterprise and offer a streamlined, scalable path for developing and deploying AI-powered enterprise applications. The high performance of Mixtral-8x7B is achieved with NVIDIA H100 Tensor Core GPUs and TensorRT-LLM, as highlighted in a recent NVIDIA Technical Blog.