NVIDIA Cosmos opens new tools to train, evaluate and deploy world models
NVIDIA expands its Cosmos platform with world-model tools: Cosmos Data Factory blueprints for curating and synthesizing training data, and Cosmos Framework skills for fine-tuning, optimizing and deploying open models. It also releases Cosmos 3 Nano simulation variants and updated inference microservices the company says run up to twice as fast.
What the article says
NVIDIA expands Cosmos, its platform of world models meant to help physical AI systems understand and simulate the real world, with new training tools, simulation checkpoints and Cosmos Data Factory blueprints. The announcement was made at GTC Berlin, and the page is dated October 21, 2026.
Cosmos Data Factory provides tools for labeling, curating, augmenting and evaluating data, with five cloud-native blueprints (autonomous vehicle dataset generation, image attribute augmentation, rare-event video generation, defect image generation, and weather and lighting augmentation). Training tools extend to more open models such as GR00T, Alpamayo, QWEN and Pi0.5; new Cosmos 3 Nano variants support simulation; and NIM microservices claim up to 2x faster inference. NVIDIA cites Porsche, ANYbotics, Ultralytics, MicroAGI, Roboflow and ADNOC as examples, and presents the package as a way to reduce physical testing and real-world data collection.
Summary written by AnnuIA in its own words from the official page of NVIDIA. The content and rights belong to the publisher: please read the original article.
More announcements from NVIDIA
- Agentic AI joins the search for extraterrestrial signals
- Thin, light and built for everything: NVIDIA RTX Spark Windows laptops gain new models, agent tools and games
- How AI is changing the pace of discovery for millions living with dementia
- Leading European financial services organizations embrace open models to build proprietary intelligence