Japanese Enterprises and Startups Leverage NVIDIA Nemotron Open Models to Build Industry-Specific AI

News Overview:

Tokyo University of Science, SoftBank Corporation, SB Intuitions Corporation, and Stockmark are adopting NVIDIA Nemotron to build proprietary AI models designed for domestic users, enterprises, and institutions in Japan, a nation facing demographic and workforce structural changes.

Avatarin, ENEOS Holdings, Hitachi, Ltd., and NTT are developing Japanese-language AI applications powered by NVIDIA Nemotron, spanning from remote-controlled robots and enterprise agents to healthcare and contact centers.

Sakana AI is integrating NVIDIA Nemotron into its model-routing platform 'Fugu,' expanding Fugu’s lineup of AI models capable of dynamically selecting the optimal model for each task.

Tokyo — July 15, 2026 — NVIDIA today announced that leading Japanese enterprises, startups, and research institutions are leveraging NVIDIA Nemotron™ open models, data, and libraries to build industry-specific AI models and applications. This accelerates the development of AI tailored to Japan’s language, industries, and workforce needs.

Open models serve as the foundation of a national AI ecosystem, enabling organizations to customize, deploy, and manage AI under their own control.

The importance of this capability is growing in Japan. As the country faces challenges such as aging populations and workforce shifts, demand is rising for AI solutions tailored to Japanese industries that enhance labor capacity, maintain productivity, and accelerate innovation.

Jensen Huang, Founder and CEO of NVIDIA, said: 'Every country and company must own and operate its own AI infrastructure. Open models make this possible. Open models give every nation, company, and researcher the freedom to validate, improve, adapt, secure, and deploy AI to meet their unique needs. NVIDIA is advancing an open AI ecosystem with Japan’s AI leaders to accelerate new discoveries, strengthen national capabilities, and ensure every society can participate in and benefit from AI innovation.'

Building Japan-Specific AI with NVIDIA Nemotron

Within Japan, developers are using NVIDIA Nemotron’s open models and datasets to build specialized AI aligned with the needs of Japanese industries and public sectors.

Tokyo University of Science has developed the open foundation model series 'Swallow' using NVIDIA Nemotron datasets and the NVIDIA NeMo™ software stack for continuous pre-training and post-training. The Swallow models retain the base model’s capabilities in English, mathematics, and coding while enhancing Japanese language processing and reasoning. Enterprises are customizing and deploying Swallow for specialized use cases such as translating financial documents and generating asset management reports.

SB Intuitions, a SoftBank subsidiary focused on generative AI research, development, and services, has trained its domestic generative AI Sarashina series using NVIDIA Nemotron, including the NVIDIA NeMo RL and Megatron-LM libraries. Sarashina3 mini has been selected by Japan’s Digital Agency for use in specific AI application cases. Additionally, SoftBank is developing and deploying the Large Telecom Model (LTM), a generative AI foundation model for the telecommunications industry, using Sarashina and NVIDIA Nemotron, aiming to achieve autonomous operation of communication networks.

Stockmark’s newly released Japanese document understanding model is based on the NVIDIA Nemotron 3 Nano Omni model. The company is also advancing development of enterprise knowledge utilization applications using NVIDIA NeMo Retriever™ and the Nemotron-Personas-Japan dataset, supporting customers in Japan’s manufacturing, energy, and chemical industries through the national project GENIAC (Generative AI Accelerator Challenge).

Transforming Japanese Industries with NVIDIA Nemotron

Japanese enterprises are leveraging NVIDIA Nemotron to modernize critical services, boost productivity, and support the workforce.

Avatarin, a startup in AI and robotics, is using NVIDIA Nemotron’s open models and NVIDIA NeMo to develop Japanese-language speech recognition and reasoning capabilities for enterprise AI agents. The NVIDIA HGX™ B300 system provides a private AI infrastructure enabling these agents to securely analyze customer conversations and access enterprise knowledge for more accurate responses. Meanwhile, NVIDIA Jetson™ powers edge AI capabilities, including digital avatar systems deployed at airports and other locations across Japan.

ENEOS Holdings is leveraging NVIDIA Nemotron open models, NVIDIA AI-Q Blueprint, and NVIDIA ALCHEMI to advance agent-based AI workflows in R&D for energy and materials. Researchers are combining technical document search, vision-language understanding, and simulation-based molecular screening to accelerate materials discovery for applications such as immersion cooling fluids and advanced catalysts.

NTT Data, a business unit of NTT, has used the NVIDIA Nemotron-Personas-Japan dataset to expand the training data for its in-house developed tsuzumi 2 model, improving question-answering accuracy and enhancing response quality for queries requiring additional knowledge. The company is exploring deployment of a scalable multi-agent framework using the NVIDIA Agent Toolkit, including NVIDIA Nemotron, to assign tasks to optimal models and achieve accurate, efficient, and autonomous enterprise workflows.

Hitachi, Ltd. is utilizing NVIDIA Nemotron and NVIDI