FrontierThe story, in brief
NVIDIA Introduces SANA-WM: A 2.6B-Parameter Open-Source World Model That Generates Minute-Scale 720p Video on a Single GPU
2.6B parameters. 60 seconds of 720p video. One GPU. NVIDIA just open-sourced a world model that changes the economics of video generation.
The KeyNews take
Why it matters
NVIDIA's SANA-WM demonstrates a major shift in multimodal AI efficiency—generating long-form video at scale on consumer hardware challenges the compute-intensive paradigm that has dominated generative video. Open-sourcing it signals NVIDIA's commitment to democratizing advanced capabilities while showcasing H100-to-consumer GPU optimization.
The key facts
7 to know2.6B-parameter model
Generates 60-second 720p video
6-DoF camera control capability
Trained on 64 H100 GPUs
Deployable on single RTX 5090
Open-source release
World model architecture (not diffusion-based)
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Researchers from NVIDIA introduce SANA-WM, an open-source camera-controlled world model that generates 60-second, 720p videos with precise 6-DoF camera control — trained on 64 H100 GPUs and deployable on a single RTX 5090. The post NVIDIA Introduces SANA-WM: A 2.6B-Parameter Open-Source World Model…