NVIDIA offers NIM (NVIDIA Inference Microservices), a set of prepackaged AI model containers that let developers and enterprises deploy large language models and other AI models in minutes rather than weeks, across clouds, data centers, and local workstations. NIM became generally available to developers in June 2024.

NIM microservices provide pre-optimized AI model containers built on inference engines including NVIDIA TensorRT, TensorRT-LLM, vLLM, and SGLang, along with industry-standard APIs, runtime dependencies, and enterprise-grade support. Developers can access NIM for models from NVIDIA and partners through the NVIDIA Developer Program for free research and development, or in production via the NVIDIA AI Enterprise software platform. NIM reduces model deployment times from weeks to minutes and has been embedded into platforms from Amazon SageMaker, Microsoft Azure AI, and over 150 ecosystem partners.