ainewsblitz.com

Breaking

NVIDIA Releases 4-Step Distilled Cosmos 3 Super Models, Claiming Up to 25x Faster Image and Video Generation

NVIDIA Releases 4-Step Distilled Cosmos 3 Super Models, Claiming Up to 25x Faster Image and Video Generation
  • Media Generation
  • Foundation Models
  • Open Source

NVIDIA has released distilled "4-step" versions of its Cosmos 3 Super world foundation models that it says generate images and video up to 25 times faster than the originals while remaining among the top-ranked open-weight systems on independent benchmarks. The new checkpoints are available now on Hugging Face and target developers building for so-called Physical AI applications such as robotics and autonomous systems.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 27,934 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year