NVIDIA Unveils Broad Suite of Open Models to Empower Industry-Specific AI Across Europe

Image: NVIDIA · Source

NVIDIA expands its open AI model portfolio with Nemotron language models, Kumo for structured business data, Cosmos for robotics, BioNeMo for protein design, and Ising for quantum computing, enabling tailored AI applications with local customization and sovereignty.

NVIDIA has announced a significant expansion of its open model portfolio designed to accelerate AI development tailored to specific industries and local languages, particularly within Europe. This move aims to provide developers, researchers, and enterprises with flexible foundation models they can adapt using proprietary data and deploy on their infrastructure, promoting sovereign AI capabilities.

Central to the expansion is the Nemotron family of open language models, which support processing text and images to help build intelligent agents capable of orchestrating complex workflows and domain-specific problem solving. The Nemotron models, including the newly introduced 3.5 Super VL and 3.5 Bolt versions, are developed in collaboration with the NVIDIA Nemotron Coalition—a global consortium of AI labs and innovators such as Black Forest Labs, H-Company, and Mistral. Key industry players like Dassault Systèmes, Salesforce, SAP, and Accenture are already leveraging Nemotron to create specialized AI agents aligned with their expertise.

Complementing Nemotron, NVIDIA has introduced the Kumo family of open foundation models designed specifically for structured enterprise data analysis. Kumo facilitates tasks such as forecasting, anomaly detection, and fraud risk scoring, reducing the need for custom machine learning pipelines. Early adopters include Coinbase, Databricks, and Revolut, with Snowflake customers utilizing the models across sectors like retail and financial services.

In the robotics and autonomous vehicle spheres, NVIDIA Cosmos and Alpamayo 2 Edge advance real-time simulation, training, and decision-making. Cosmos provides world models and data curation tools to improve robot perception and interaction in physical environments, while Alpamayo 2 Edge integrates Nemotron-based real-time reasoning optimized for NVIDIA’s DRIVE AGX Thor platform, targeting efficient deployment on autonomous vehicle hardware.

The BioNeMo platform furthers NVIDIA’s commitment to life sciences by offering an open protein design model, RFD4-Proteina, developed alongside the University of Washington’s Baker Lab. This model expedites protein sequence and structure generation to assist scientific discovery, delivering up to 13 times the speed of comparable approaches in benchmarks.

Addressing quantum computing, NVIDIA Ising offers versatile open models to enhance quantum processor calibration and error correction across varied qubit technologies. The model is supported by partnerships with companies like IQM Quantum Computers and is integrated into AWS’s Parallel Computing Service, enabling broader access for researchers.

All these models and tools are openly accessible via platforms including Hugging Face, GitHub, and NVIDIA’s own services, often packaged as NVIDIA inference microservices for streamlined deployment. This open approach is intended to foster innovation, transparency, and sovereignty in AI development by allowing organizations worldwide to customize models for language, cultural, and domain-specific needs.

By uniting global expertise through initiatives like the NVIDIA Nemotron Coalition and releasing a broad suite of interoperable AI foundation models, NVIDIA aims to set a new standard for collaborative, specialized AI ecosystems extending from business intelligence to scientific research, robotics, and quantum technologies.