Model
Inference & Serving
Mochi 1
1Open-source 2model 3generates 4high-fidelity 5video 6motion
the six Ws · specification
W1
Who
Genmo AI released Mochi 1.
W2
What
Mochi 1 is a 10 billion parameter open-weight video diffusion model built to produce high-fidelity motion and strong prompt adherence.
W3
Where
Hosted on GitHub at genmoai/mochi and on Hugging Face under genmo.
W4
When
Released in October 2024.
W5
Why
At release it was billed as the largest openly licensed video generation model, aiming to close the gap with closed commercial systems.
W6
With
Built on PyTorch with a custom AsymmDiT architecture, runnable via the ComfyUI or Diffusers integrations.
W7
Watch
Released under the Apache 2.0 license by Genmo; fully open weights with no gating.
alternatives
works with