NVIDIA's Cosmos3 4-Step Distills Top Open-Weight Image-to-Video Leaderboards
ArtificialAnlys · x · 2026-07-31
NVIDIA released 4-step distilled variants of its 64B Cosmos3-Super world foundation model for text-to-image and image-to-video generation.
Key highlights:
- Massive Step Reduction: Cuts text-to-image inference from 50 steps to just 4, and image-to-video from 35 to 4, removing the need for classifier-free guidance.
- Leaderboard Success: The image-to-video model is the new #1 open-weights model on the Artificial Analysis Arena (15th overall), while text-to-image ranks #3 among open weights (25th overall).
- Open & Commercial: Both models are available on Hugging Face under the OpenMDW-1.1 license, permitting commercial use.
- Targeted Use Cases: The Cosmos 3 family is positioned for physical AI (robotics, autonomous vehicles), outputting up to 720p and 8-second video clips by default.
More from Models
- Mistral touts STEM benchmark win, but leaderboard shows it ranking sixth overall — MistralAI · 2026-10-16
- Local tool calling compared: how Gemma 4, Llama 3 and Mistral differ — leslysandra · 2026-10-16
- $2 Gemini Pro accounts flood the gray market, spawning waves of AI spam comments — huangyun_122 · 2026-10-16
- Haider: Opus 5.5 vs Astra on SWE tasks — Astra has a higher floor, reasoning in a tier of its own — haider1 · 2026-10-16
- Artificial Analysis launches Decision Index for the emerging decision model category — ArtificialAnlys · 2026-10-16
- Liquid AI's d1-ultra tops Artificial Analysis Decision Index with 85% non-hallucination rate — ArtificialAnlys · 2026-10-16