self-hosted/ai
§01·model · /models

Mochi 1

videoactiveApache-2.0

10B-parameter open text-to-video diffusion model by Genmo. Runs via diffusers MochiPipeline in bf16 (~20 GB transformer weights); high-fidelity motion and strong prompt adherence. Released under Apache-2.0.

§02·GPUs that run this model
4 total
GPUVRAMSeriesBest speedMin VRAMWorksBenchmarksRecipe
RTX 309024GB30~0recipecheck ↗
RTX 3090 Ti24GB30~0recipecheck ↗
RTX 409024GB40~0recipecheck ↗
RTX 509032GB50~0recipecheck ↗

benchmarked·~ runs via recipe (not benchmarked)· untested·doesn't fit