NVIDIA distills Cosmos3 Super image-to-video to 4 steps with a 64B model
multimodalart · x · 2026-07-27
NVIDIA says it distilled Cosmos3 Super Image-to-Video to run in just 4 steps.
- The model is currently the top open-weights image-to-video system on the Artificial Analysis leaderboard.
- The catch is size: it has 64B parameters.
- Despite that, it reportedly runs on Hugging Face Spaces, and the 4-step version is fast enough to be practically usable, so NVIDIA is encouraging people to try it.
More from Models
- Kimi K3 launches on SGLang with 423 tok/s and 11 cloud partners — ying11231 · 2026-07-28
- Ollama Adds Kimi K3: 1M Context Window and Native Vision Support — ollama · 2026-07-27
- Kimi K3 reportedly improves training efficiency by 2.5× — zephyr_z9 · 2026-07-27
- Kimi K3 goes live on Modal with custom DFlash speculative decoding for lossless speedup — AAAzzam · 2026-07-27
- Kimi K3 with 2.8T parameters and 1M context now supported on vLLM — ricklamers · 2026-07-27
- Kimi K3 could become a cheap distillation base for personal, local models — victormustar · 2026-07-27