NVIDIA Cosmos Opens New Tools to Train, Evaluate and Deploy World Models
Web page · https://blogs.nvidia.com/blog/cosmos-open-train-evaluate-deploy-world-models
To help developers rapidly build world models, which help physical AI systems understand and simulate the real world, NVIDIA has expanded the Cosmos platform with new training tools, simulation checkpoints and Cosmos Data Factory blueprints. Ranking top five on six different open models leaderboards, the NVIDIA Cosmos platform gives world model builders the frontier open […]
First seen by Oriel on .
AI summary
NVIDIA expanded its open Cosmos platform with data-generation blueprints, model training and optimization tools, simulation checkpoints, and faster deployment options for physical AI developers.
- Cosmos Data Factory offers five cloud-native blueprints for autonomous-vehicle datasets, image augmentation, event videos, defect images, and video augmentation.
- Porsche expects its shared AV data flywheel to reduce repeated test drives by up to 90% and development cycle times by up to 85%.
- ANYbotics generated over 1,500 synthetic video clips to train inspection robots, while Ultralytics improved defect localization and recall using targeted synthetic images.
- Cosmos Framework supports fine-tuning, reinforcement learning, distillation, and quantization across Cosmos 3 and compatible open models; MicroAGI distilled a 32-billion-parameter reasoner to an 8-billion-parameter model with no accuracy loss.
- Updated Cosmos 3 NIM microservices deliver up to 2x faster inference with no quality loss in validated configurations, and Metropolis VSS 3.3 monitors 40% more video streams on the same GPU.
#world models #physical ai #synthetic data #video analytics
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| NVIDIA Blog | https://blogs.nvidia.com/?p=99128 | https://blogs.nvidia.com/blog/cosmos-open-train-evaluate-deploy-world-models/ Read discussion |