NVIDIA today announced the NVIDIA AI Foundry service and the NVIDIA NIM inference microservice, which power generative AI for enterprises around the world with Meta’s open-source AI model collection, Lamar 3.1.

NVIDIA AI Foundry enables companies to create custom “supermodels” tailored to domain-specific industry use cases using Lama 3.1 and NVIDIA solutions. Companies can train these supermodels using their own data, as well as synthetic data generated from Lama 3.1 405B and NVIDIA Nemotron reward models.

NVIDIA AI Foundry is powered by the NVIDIA DGX Cloud AI platform, giving enterprises massive compute resources that can easily scale as their AI demands change.

The new service will support the construction of custom large language models (LLMs) for enterprises as well as countries developing their own AI strategies. Countries in each language-speaking region have a need to build customized LLMs with domain-specific knowledge for generative AI applications that reflect their unique business or culture.

To power enterprise deployments of the current production Llama 3.1 model for AI, we are now releasing the NVIDIA NIM inference microservice for Llama 3.1 models on ai.nvidia.com. The NIM microservice is the fastest way to deploy Llama 3.1 models in production, and is claimed to deliver up to 2.5x higher throughput than running inference without NIM.

Enterprises can combine the Llama 3.1 NIM microservice with the new NVIDIA NeMo Retriever NIM microservice to build state-of-the-art search pipelines for AI copilots, assistants, and digital human avatars.

Accenture, a global professional services firm, is now the first to adopt NVIDIA AI Foundry for its clients looking to deploy generative AI applications that reflect their own use cases, culture, language and industry, building a custom Llama 3.1 model using Accenture’s AI Refinery framework.

Companies in healthcare, energy, financial services, retail, transportation, and telecommunications are already using NVIDIA NIM microservices for Llama. Among the first enterprises to adopt the new NIM microservices for Rama 3.1 are Aramco, AT&T, and Uber.

“Meta’s open-source model, Rama 3.1, is a significant moment for the adoption of generative AI in enterprises around the world,” said Jensen Huang, CEO. “NVIDIA AI Foundry has fully integrated Rama 3.1 and stands ready to help enterprises build and deploy custom Rama supermodels.”