Introducing Cosmos 3 Edge
Quick Answer
NVIDIA has launched Cosmos 3 Edge, a 4-billion-parameter model for real-time reasoning and action generation in physical AI systems, achieving top performance in vision analytics and robot policy learning.
Quick Take
It operates efficiently on NVIDIA edge devices, delivering 32 actions per inference at 15 Hz.
Key Points
- Cosmos 3 Edge is optimized for NVIDIA RTX PRO and Jetson devices.
- Ranks #1 on VANTAGE-Bench for vision analytics among 4B parameter models.
- Utilizes two transformer towers for multimodal understanding and action prediction.
- Supports real-time control with 640x360 resolution observations.
- Includes a post-trained robot manipulation policy for pick-and-place tasks.
DeepSignal Analysis
What happened
NVIDIA has introduced Cosmos 3 Edge, a 4-billion-parameter model designed for real-time reasoning and action generation in physical AI systems. It operates efficiently on NVIDIA edge devices, achieving 32 actions per inference at a frequency of 15 Hz. The model ranks first on VANTAGE-Bench for vision analytics among similar-sized models.
Key evidence
- Cosmos 3 Edge is a 4-billion-parameter model that helps robots and vision AI agents understand their surroundings and generate actions in real time.
- The model achieves real-time control at 15 Hz and generates 32 actions per inference on NVIDIA Jetson Thor.
- Cosmos 3 Edge ranks #1 on VANTAGE-Bench for vision analytics and is state-of-the-art for robot policy learning among models of similar size.
Why it matters
The launch of Cosmos 3 Edge signifies a step forward in the capabilities of physical AI systems, enabling them to operate effectively in memory-constrained environments like factories and hospitals. Its high throughput and efficiency on edge devices could enhance the deployment of AI in real-world applications, particularly in robotics and smart infrastructure. This model's performance metrics may influence future developments in AI model design and deployment strategies.
Source Excerpt
A Blog post by NVIDIA on Hugging Face
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from Hugging Face
See more →
From Hugging Face to Amazon SageMaker Studio in one click
Hugging Face has launched a deep-link integration with Amazon SageMaker Studio, allowing developers to seamlessly transition from model discovery to deployment with a single click. This integration streamlines the process by pre-configuring permissions and providing GPU quota visibility, significantly reducing the time from model selection to experimentation.
