News summary:
- Institution of Science Tokyo, SoftBank Corp., SB Intuitions Corp. and Stockmark are using NVIDIA Nemotron to develop locally developed AI models to serve Japanese users, companies and institutions amid the country’s changing demographics and workforce.
- Japanese companies Avatarin, ENEOS Holdings, Hitachi and NTT DATA are using NVIDIA Nemotron to develop Japanese-language AI applications, from remote presence robotics to corporate agents and specialized medical and contact centers.
- Sakana AI integrates NVIDIA Nemotron into its Fugu model routing platform, expanding the set of AI models that Fugu can intelligently orchestrate to dynamically select the best model for each task.
NVIDIA today announced that leading Japanese companies, startups and research institutions are developing industry-specific AI models and applications using NVIDIA Nemotron™ open models, data and libraries, accelerating the development of AI tailored to Japan’s language, industries and workforce.
Open models form the foundation of national AI ecosystems, giving companies the ability to adapt, deploy and govern the AI they control.
In Japan, these skills are becoming increasingly important as the country grapples with an aging population and changing workforce, driving demand for AI tailored to local industries that helps empower the workforce, maintain productivity and accelerate innovation.
“Every nation and every company should own and control their intelligence infrastructure. Open models make that possible,” said Jensen Huang, founder and CEO of NVIDIA. “They give countries, companies and researchers the freedom to study, improve, adapt, secure and deploy AI for their own needs. Together with Japan’s AI leaders, we are advancing an open AI ecosystem that accelerates discovery, strengthens national capabilities and ensures that every society can participate in and benefit from the AI revolution.”
Building specialized AI for Japan with NVIDIA Nemotron
Across Japan, developers are developing specialized AI using open models and datasets from NVIDIA Nemotron, adapting them to the country’s industries and public sector needs.
The Institute of Science Tokyo has developed its Swallow family of open foundation models using NVIDIA Nemotron datasets and the NVIDIA NeMo™ software stack for continuous pre- and post-training. Swallow models improve Japanese language and reasoning performance while preserving the underlying models’ core competencies in English, mathematics, and coding. Companies customize and implement Swallow for specific use cases, including translating financial documents and generating asset management reports.
SB Intuitions Corp., the generative AI research subsidiary of SoftBank Corp., trained its Sarashina series of homegrown generative AI models using NVIDIA Nemotron, including the NVIDIA NeMo RL and Megatron LM libraries. Sarashina3 mini was selected by the Japanese digital agency for use in specific AI use cases. SoftBank Corp. has also developed and deployed a Large Telecom Model (LTM) using both Sarashina and NVIDIA Nemotron to enable autonomous telecom network operations.
Stockmark’s dedicated document understanding model for Japanese languages, released today, is based on the NVIDIA Nemotron 3 Nano Omni model. The company also develops enterprise knowledge applications using NVIDIA NeMo Retriever™ and the Nemotron-Personas-Japan dataset, and serves customers in the Japanese manufacturing, energy and chemical industries through the Japanese national Generative AI Accelerator Challenge project.
Transforming Japan’s industry with NVIDIA Nemotron
Japanese companies use NVIDIA Nemotron to modernize critical services, improve productivity and support the country’s workforce.
AI and robotics startup Avatarin uses the open models NVIDIA Nemotron and NVIDIA NeMo to develop Japanese-language speech and reasoning capabilities for enterprise AI agents. NVIDIA HGX™ B300 systems provide the private AI infrastructure that enables these agents to securely analyze customer conversations and access corporate knowledge for more accurate answers, while NVIDIA Jetson™ Edge AI capabilities, including digital avatar systems, deployed at airports and other locations across Japan.
ENEOS Holdings leverages NVIDIA Nemotron open models with NVIDIA AI-Q Blueprint and NVIDIA ALCHEMI NIM microservices to power agent AI workflows for energy and materials research and development. Researchers are using these technologies to integrate technical document search, image and language understanding, and simulation-based molecular screening, helping to accelerate materials research for applications such as immersion cooling fluids and advanced catalysts.
NTT DATA, an operating subsidiary of NTT, used NVIDIA Nemotron-Personas-Japan to augment training data for its proprietary Tsuzumi-2 model, improving accuracy in answering questions and improving answers to questions that require additional knowledge. The company also aims to deliver a scalable multi-agent framework that leverages the NVIDIA Agent Toolkit, including NVIDIA Nemotron, to route tasks to the best models and drive accurate, efficient and autonomous enterprise workflows.
Hitachi develops physical AI technologies to address real-world operational challenges by leveraging the open NVIDIA Nemotron and NVIDIA Cosmos™ models, as well as its proprietary information technology (IT) and operational technology (OT) expertise. As part of a multi-agent orchestration platform, these technologies are designed to connect and coordinate IT and OT operations and help transform enterprise-wide business processes across complex workflows.
Sakana AI is working with NVIDIA to integrate NVIDIA Nemotron into its Fugu model orchestration platform, expanding the range of AI models that Fugu can intelligently orchestrate to dynamically select the best model for each task in agent AI workflows. By routing each request to the model best suited for the task, Fugu helps developers balance accuracy, performance, and cost across multiple open and proprietary AI models. Fugu shows how thoughtful orchestration can unlock capabilities beyond what a single model achieves alone. Early performance results on complex, real-world coding tasks support the promise of coordination as a path to more powerful AI.
Open and customizable, can be used anywhere
Nemotron models are published with open weights, datasets, and recipes, giving organizations the visibility and control to customize models for domain-specific workflows and deploy them wherever their applications and data reside.
Developers can use NVIDIA NeMo to customize, evaluate, and optimize models for their use cases and deploy them in environments that meet regulatory, government, and data localization requirements.
Nemotron models are available on Hugging Face, ModelScope, OpenRouter and build.nvidia.com as NVIDIA NIM™ microservices, as well as through NVIDIA Cloud partners, inference platforms and cloud service providers.
https://nvidianews.nvidia.com/news/japans-enterprises-and-startups-build-industry-specialized-ai-with-nvidia-nemotron-open-models
