NVIDIA Open Models Advance Europe’s Sovereign Intelligence Across Industries

NVIDIA unveiled a broadened suite of open models spanning language, robotics, biology, and quantum computing, enabling developers and enterprises to tailor applications to local languages and sector-specific needs. The portfolio includes Nemotron 3.5 Super VL, Nemotron 3.5 Bolt, Nemotron Retriever models, Kumo family models, Cosmos 3.5 world models, Alpamayo 2 Edge, BioNeMo RFD4-Proteina, and Ising models, all designed for customization and deployment on controlled infrastructure.

Nemotron models, developed with partners like Black Forest Labs, H-Company, and Mistral, support text and image processing to help agents orchestrate workflows and solve specialized tasks at reduced costs. Enterprises such as Dassault Systèmes, SAP, and WPP are collaborating with Nemotron to build domain-specific agents, supported by a post-training ecosystem including Scale AI and Google Cloud.

Kumo open foundation models and libraries enable forecasting, fraud risk scoring, and action recommendations from structured business data, reducing the need for separate machine learning pipelines. Early adopters like Cognite, Coinbase, and Snowflake are using Kumo to accelerate AI development in sectors including advertising, energy, and financial services.

Cosmos 3.5 world models introduce new post-training tools and simulation checkpoints to improve robotics and autonomous vehicle development, while BioNeMo RFD4-Proteina accelerates protein design by generating sequences and 3D structures 13 times faster than prior benchmarks. Ising models support quantum computing advancements, with IQM Quantum Computers and SEEQC leveraging them for calibration and error correction.