NVIDIA Launches Regional AI Models to Support Sovereign AI Efforts

Tokyo, Japan / Taipei, Taiwan - NVIDIA (NASDAQ: NVDA) announced the release of four new AI microservices on Wednesday to support sovereign AI initiatives worldwide. 

The NVIDIA NIM microservices, designed to work with regional language models, are expected to enhance the development and deployment of generative AI applications tailored to local needs.

The new offerings include models trained on Japanese and Mandarin data, such as Llama-3-Swallow-70B and Llama-3-Taiwan-70B, as well as the RakutenAI 7B models for English and Japanese. 

These models are optimized for improved regional language understanding, legal tasks, and translation performance.

NVIDIA's microservices, available with NVIDIA AI Enterprise, are optimized for inference using the company's TensorRT-LLM open-source library. 

The company claims these services can provide up to 5x higher throughput for Llama 3 70B models, potentially lowering production costs and improving user experiences.

Several organizations, including the Tokyo Institute of Technology, Chang Gung Memorial Hospital in Taiwan, and various Taiwanese companies in different sectors, are already adopting these microservices.

NVIDIA also highlighted its AI Foundry platform, which allows enterprises to fine-tune models for specific business needs and create custom NIM microservices.

The company anticipates that these tools will accelerate the deployment of AI applications across industries such as healthcare, finance, manufacturing, education, and legal services.