NVIDIA expands Cosmos with world model training tools, simulation checkpoints and Data Factory blueprints
NVIDIA has expanded the Cosmos platform with new training tools, simulation checkpoints and Cosmos Data Factory blueprints. The company says Cosmos ranks in the top five on six different open model leaderboards.
NVIDIA has expanded its Cosmos platform with new training tools, simulation checkpoints and Cosmos Data Factory blueprints, according to a post on the NVIDIA blog.
The additions are aimed at developers building world models, which help physical AI systems understand and simulate the real world. The post is titled "NVIDIA Cosmos Opens New Tools to Train, Evaluate and Deploy World Models."
What was announced
- New training tools for world model builders.
- Simulation checkpoints added to the platform.
- Cosmos Data Factory blueprints, also new in this expansion.
NVIDIA says the Cosmos platform ranks in the top five on six different open model leaderboards. This is a company claim. It describes the platform as giving world model builders frontier open models.
Parameter counts, context windows, pricing and benchmark scores were not part of the announcement summary.
What this means
The expansion covers training, evaluation and deployment of world models under the Cosmos platform, according to NVIDIA's own framing. Developers working on physical AI systems get additional tooling, checkpoints and data blueprints in one platform rather than needing to assemble them separately. The leaderboard standing is NVIDIA's own claim and should be checked against the leaderboards themselves.
Related Articles
NVIDIA Cosmos3-H-Dreams runs real-time generative cholecystectomy simulation on a single RTX PRO 6000
NVIDIA released Cosmos3-H-Dreams, a world foundation model that generates action-conditioned surgical video in real time on a single RTX PRO 6000 GPU. It is trained on clinical cholecystectomy recordings from the CMR Surgical Versius platform and ships in four-step and two-step distilled checkpoints. NVIDIA claims it is the first real-time generative surgical simulator trained on clinical cholecystectomy data with live kinematic control.
NVIDIA Kumo Relational predicts from databases using up to 10,000 in-context examples, no training
NVIDIA released Kumo Relational, a pre-trained foundation model that predicts outcomes such as churn, demand, and fraud directly on relational databases in a single forward pass. NVIDIA claims it beats supervised relational deep learning and expert-engineered gradient-boosted trees across 41 tasks on 15 databases, and surpasses AutoGluon ensembles by 8% MRR on SAP SALT.
NVIDIA releases Kumo Forecast, an Apache 2.0 time-series model built on a ~340M-parameter MOMENT-1-large encoder
NVIDIA has released Kumo Forecast (formerly NV-Tesseract Forecasting), a multivariate time-series forecasting model under Apache 2.0. It pairs a frozen MOMENT-1-large encoder of roughly 340 million parameters with a trainable forecasting head, optional nearest-neighbor retrieval over historical data (DARR), and a built-in interpretability framework.
NVIDIA opens RFD4-Proteina protein design model on Hugging Face Spaces, free dedicated GPUs until Nov 4
NVIDIA is sponsoring free access to a dedicated Hugging Face Space for RFD4-Proteina, an open BioNeMo protein design model co-developed with the University of Washington's Institute for Protein Design. Access runs from October 21 to November 4, 2026. NVIDIA claims the model produced 76.40 qualifying designs per GPU-hour at high thinking, versus 13.24 for the next-best method.
Comments
Loading...