FA

FastH3 GGUFs

FastvideoGGUF by realrebelaiUpdated 2026-08-31
Download on Hugging Face
Parameters35B
Context—
Layers—
Architecturewan
Smallest quant19.8 GB
Quick answer

FastH3 GGUFs needs about 22.7 GB of memory at Q4_K_M (a 19.8 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 2 quants, from 19.8 GB (Q4_K_M) to 24.2 GB (Q5_K_M).

FastH3 4‑step video model distilled to 35B parameters

FastH3 4‑step is a 35B video model distilled from FastVideo/FastVideo‑FastH3‑4‑step‑Preview‑v1‑VSA‑DataFree. It supports text‑to‑audio‑video generation with four transformer forwards. The GGUF files are quantized by realrebelai and require VSA attention support.

Good at

  • Text‑to‑audio‑video generation
  • Four‑step inference
  • Data‑free distilled weights

Limits

  • Requires VSA attention code; stock ComfyUI will fail
  • Large file sizes; pick by system RAM
  • No vision support beyond video generation

Best for

  • Creative video production
  • Rapid prototyping of video models
  • Low‑cost inference on consumer GPUs

Who it’s for. Developers and researchers needing a 35B video model that fits on consumer hardware with proper VSA support. Suitable for those who can manage the required ComfyUI branches and CUDA kernels.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_K_M19.8 GB 22.7 GB
Too big Get file →
Q5_K_M24.2 GB 27.6 GB
Too big Get file →

For 30‑GB RAM, use Q4_K_M (19.84 GB, ~22.7 GB memory at 8k context). For 40‑GB RAM, use Q5_K_M (24.21 GB, ~27.6 GB memory at 8k context). Estimates are approximate.

What GPU do you need for FastH3 GGUFs?

These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: no quant fits at an 8K context.
  • 12 GB VRAM: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: the best quant that fits is Q4_K_M (19.8 GB file, about 22.7 GB in use).
  • 32 GB VRAM: the best quant that fits is Q5_K_M (24.2 GB file, about 27.6 GB in use).

Which FastH3 GGUFs quant should you choose?

Q4_K_M is the usual starting point for FastH3 GGUFs: it balances file size and quality well. Here it is a 19.8 GB file.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturewan
Parameters35B

Source

Quantized byrealrebelai
FamilyFastvideo
Licenseother
Downloads (HF)697
Last updated2026-08-31

Frequently asked questions about FastH3 GGUFs

How much VRAM does FastH3 GGUFs need?

About 22.7 GB at Q4_K_M with an 8K context window. That is the 19.8 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run FastH3 GGUFs on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 22.7 GB, so you would need a larger GPU or CPU offloading.

What is the file size of FastH3 GGUFs GGUF?

It depends on the quant: Q4_K_M is 19.8 GB, Q5_K_M is 24.2 GB.

Where can I download FastH3 GGUFs GGUF?

The files are on Hugging Face in the realrebelai/FastH3_GGUFs repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does FastH3 GGUFs use?

Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run FastH3 on 8 GB VRAM?

No. The model requires 35B parameters and the GGUF files exceed 19 GB, so 8 GB VRAM is insufficient.

Do I need a special ComfyUI branch?

Yes. The model needs the vsa branch and compiled sparse attention kernels to function correctly.

What license applies to FastH3?

The model inherits MiniMax‑H3 Community License terms from the parent model.

Similar models

Browse all →

More from realrebelai

All by realrebelai →

More Fastvideo models

All Fastvideo →

Other large models

All large models →