Regarding Meta’s earlier announcement that it can handle 4050 billion sets of parametersLarge Natural Language Model Llama 3.1NVIDIA announced earlier that Llama 3.1 was trained using more than 16000 NVIDIA H100 accelerators and a dataset of more than 15 trillion tokens. It also enables businesses to build customized artificial intelligence applications through the NVIDIA AI Foundry service and NVIDIA NIM microservices.
NVIDIA stated that its NVIDIA AI Foundry service is built on the NVIDIA DGX Cloud AI platform and is scalable to meet the needs of AI operations, such as scale. Furthermore, it leverages the NVIDIA NIM microservice to rapidly deploy AI model applications, enabling the rapid construction of various AI-driven network services.
Meta's newly released large-scale natural language model, Llama 3.1, combined with NVIDIA's software and computing technologies, enables the construction of customized "super models" for specific application needs. For example, synthetic data generated by the NVIDIA Nemotron Reward model can be used to train different customized "super models."
Enterprises can now download the NVIDIA NIM microservices for Llama 3.1 from the NVIDIA website to build applications including AI digital assistants and digital avatars. Global professional services firm Accenture is already a pioneer in adopting NVIDIA AI Foundry services, using its Accenture AI Refinery framework to build customized Llama 3.1 application models, thereby accelerating the development of AI technology applications.
Businesses that need additional training data to create domain-specific model applications can use the 3.1 billion parameter version of Llama 4050 alongside synthetic data generated by the 4 billion parameter version of NVIDIA Nemotron-3400 to improve the accuracy of their customized "super models." Businesses with their own training data can create customized Llama 3.1 models through the NVIDIA NeMo service.
In addition, NVIDIA and Meta are collaborating to provide a streamlined solution for Llama 3.1, allowing developers to build smaller, customized Llama 3.1 models, allowing businesses to deploy small-scale, automatically generated artificial intelligence technology on workstations or laptops.
Companies in healthcare, financial services, retail, transportation, and telecommunications are already using NVIDIA NIM microservices. The first wave of companies to adopt Llama 3.1 and the new NVIDIA NIM microservices include Saudi Aramco, AT&T, Uber, and others, enabling faster execution and responsiveness.
