Tier 3 Practical open-weight
What the publisher claims: open-source, open weights
The one-liner: An Apache-2.0 video model whose practical limits are technical, not legal.
The six questions
Model weights are downloadable from the publisher's Hugging Face repository without gating.
Apache-2.0 on the published checkpoints, an OSI-approved permissive license with no revenue or use threshold.
Self-hostable with published inference code, though the model is limited to 480p and roughly six-second silent clips.
The technical blog and paper describe the asymmetric diffusion transformer design; the training corpus and recipe are not published.
Limited.
Genmo released Mochi under Apache 2.0 with a public inference repository that accepts outside pull requests, but the model was developed by a startup whose direction no outside contributor could influence.
License
Apache-2.0 OSI-approved — license source ↗
- Standard Apache-2.0 obligations only
What is actually released
- weights (preview and release checkpoints)
- inference code
- technical report and release blog
Why it matters
Useful because the reasons to use it are hardware and resolution rather than licence terms — openness is not the binding constraint here.
Sources
- https://huggingface.co/genmo/mochi-1-preview
- https://www.genmo.ai/blog/mochi-1-a-new-sota-in-open-text-to-video
- https://siliconangle.com/2024/10/22/genmo-introduces-mochi-1-open-source-text-video-generation-model
Derived from the openness dataset behind But Is It Open? — snapshot 2026-11-07 ↗. Scores are qualitative readings of the cited evidence, not verdicts; tier measures openness, not intelligence.