MiniMax H3

MiniMax H3 is an audio-video generation family from MiniMax. Mold can discover, download, verify, repair, inventory, and remove two compact Comfy variants. The files are downloaded directly from their pinned Hugging Face repositories; Mold does not bundle or mirror the weights.

CUDA FL2VA is the first supported runtime

Both compact variants can be downloaded on any Mold host. Mold's SM89 CUDA release can run the compact FL2VA model for the supported request profile below. Ref2VA execution and the Metal and CPU backends remain unavailable. Broader request shapes also remain unavailable until those paths are implemented and tested; Mold reports that limitation normally rather than treating it as a licensing or authorization failure.

Compact variants

Model	Task	Total pull	Runtime status
minimax-h3-fl2va:comfy-pruned-int8	First/last-frame conditioning with audio	42.482 GB	CUDA generation; first-frame profile
minimax-h3-ref2va:comfy-pruned-int8	Reference media to video with audio	42.482 GB	Downloadable; execution unavailable

Pull a variant from the CLI, or install it from Models → Discover in Mold Studio:

mold pull minimax-h3-fl2va:comfy-pruned-int8
mold pull minimax-h3-ref2va:comfy-pruned-int8

The files are revision-pinned and SHA-256 verified before Mold marks the model complete. Raw repository IDs, custom manifests, configured aliases, and live catalog recipes cannot substitute for either registered graph.

Download size and sources

Each compact variant has the same component graph except for its task-specific transformer:

Component	Bytes	Decimal size	Upstream source
Task transformer	20,970,379,616	20.970 GB	Comfy-Org/MiniMax-H3
Qwen3-VL NVFP4-AWQ text encoder	15,687,142,551	15.687 GB	Comfy-Org/MiniMax-H3
FP16 video VAE	5,207,808,496	5.208 GB	Comfy-Org/MiniMax-H3
FP32 audio VAE	605,254,808	0.605 GB	Comfy-Org/MiniMax-H3
Tokenizer, processor, scheduler, and component configs	11,504,847	0.012 GB	MiniMaxAI/MiniMax-H3
One complete variant	42,482,090,318	42.482 GB	Both pinned repositories

The encoder, VAEs, and common support files are shared between the variants. After one complete variant is installed, adding the other downloads its 20.970 GB transformer and 546-byte task config. Both variants together occupy 63.452 GB (63,452,470,480 bytes) of model payloads, excluding filesystem and Hugging Face cache overhead.

Sizes above are decimal gigabytes (1 GB = 1,000,000,000 bytes) and describe downloads and disk use, not peak VRAM. They come from Mold's registered, full-file manifest identities rather than estimates from repository listings.

Supported FL2VA request

The initial compact CUDA implementation supports this request profile:

• an SM89 CUDA GPU with sufficient VRAM and the H3 attention/runtime operators enabled
• 1344x768, batch size 1
• exactly 124 frames at 24 fps
• exactly 21 terminal-inclusive sampler grid points (20 model evaluations)
• one required first-frame image and no last-frame endpoint
• MP4 output with synchronized generated audio

Mold rejects rather than silently resizing, rerouting, changing steps, dropping the source image, or falling back to another backend. A downloaded checkpoint can remain stored on an unsupported host; Create and request routing become available only when that host advertises the matching CUDA runtime capability. The public SM89 runtime uses a source-controlled conservative memory profile; it does not require private authorization or qualification-record files.

License and support boundary

H3 model weights and upstream model assets use the MiniMax H3 Community License, at pinned revision bfc8ed0353f5a9733be73e6b2c98ec0948195b86, not Mold's MIT license. Review those terms for your intended use. Mold's source code and H3 integration remain under Mold's repository license. Mold downloads the weights directly from the pinned upstream repositories and verifies every file; it does not bundle or mirror the payloads in Mold releases. The project's license and integration record documents the completed governance decision.

The completed project review authorizes H3 use in every territory and across Mold's CLI, server/API, Discord, desktop, web, iPhone, TUI, gallery, remote-client, shared-server, and hosted paths. It also covers generated-output distribution and model distribution or redistribution. Technical availability remains limited to routes Mold has implemented and qualified; authorization does not make an unsupported task, device, or request shape runnable.

The license link and notice in this guide and Mold's README are the project's required user-facing license, attribution, disclosure, downstream-term, and acceptable-use delivery. Mold does not require a separate clickthrough, geolocation check, H3-specific generated-content label, downstream contract, or surface-specific acceptable-use control. Existing Mold authentication, validation, capability, safety, and operational controls continue to apply.

The official BF16 checkpoints remain hidden qualification references. Their much larger artifact graphs are not public Mold download options.