MI

Sakura MiniMax H3

MinimaxGGUF by webmp3Updated 2026-10-13
Download on Hugging Face
Parameters20.1B
Context—
Layers—
Architecture—
Smallest quant12.6 GB
Quick answer

Sakura MiniMax H3 needs about 14.6 GB of memory at Q4_K_M (a 12.6 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 4 quants, from 12.6 GB (Q4_K_M) to 21.4 GB (Q8_0).

Sakura MiniMax H3 – 21B parameter video‑audio generation model

Sakura MiniMax H3 is a 21‑billion‑parameter model derived from MiniMax‑H3. It supports multimodal input and can generate video with synchronized stereo audio. The model is available in several GGUF quantizations for use with llama.cpp.

Good at

  • Multimodal text‑image‑audio understanding
  • Video generation up to 2K resolution
  • Audio‑video synchronization

Limits

  • No vision support in this GGUF version
  • Requires large VRAM for higher quantizations
  • License is other, not specified

Best for

  • Generating short video clips with audio
  • Creating multimodal content for creative projects
  • Testing video‑audio generation pipelines

Who it’s for. Developers and researchers who need a 21B parameter model for video‑audio generation. The model is suitable for those who can allocate sufficient VRAM and are comfortable with community‑quantized GGUF files.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_K_M12.6 GB 14.6 GB
Too big Get file →
Q5_K_M14.8 GB 17.1 GB
Too big Get file →
Q6_K16.6 GB 19.0 GB
Too big Get file →
Q8_021.4 GB 24.5 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (12.58 GB file, 14.6 GB estimated memory). For 12 GB VRAM, Q5_K_M (14.79 GB file, 17.1 GB estimated memory). For 16 GB VRAM, Q6_K (16.55 GB file, 19 GB estimated memory). Higher quantizations require more VRAM and may fail to load.

What GPU do you need for Sakura MiniMax H3?

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: the best quant that fits is Q4_K_M (12.6 GB file, about 14.6 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K (16.6 GB file, about 19.0 GB in use).
  • 32 GB VRAM: the best quant that fits is Q8_0 (21.4 GB file, about 24.5 GB in use).

Which Sakura MiniMax H3 quant should you choose?

Q4_K_M is the usual starting point for Sakura MiniMax H3: it balances file size and quality well. Here it is a 12.6 GB file.

The Q8_0 file is 21.4 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Parameters20.1B

Source

Quantized bywebmp3
FamilyMinimax
Licenseother
Downloads (HF)20,016
Last updated2026-10-13

Frequently asked questions about Sakura MiniMax H3

How much VRAM does Sakura MiniMax H3 need?

About 14.6 GB at Q4_K_M with an 8K context window. That is the 12.6 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 Sakura MiniMax H3 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 14.6 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Sakura MiniMax H3 GGUF?

It depends on the quant: Q4_K_M is 12.6 GB, Q5_K_M is 14.8 GB, Q6_K is 16.6 GB, Q8_0 is 21.4 GB.

Where can I download Sakura MiniMax H3 GGUF?

The files are on Hugging Face in the webmp3/Sakura-MiniMax-H3-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Sakura MiniMax H3 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 Sakura MiniMax H3 on 8 GB VRAM?

Use the Q4_K_M quantization; it requires about 14.6 GB of VRAM at 8k context, so it will not fit on 8 GB. You would need a higher VRAM GPU.

Does the model support image input?

The GGUF version focuses on video‑audio generation; image input is not supported in this quantized release.

What is the context length of this model?

The trained context tokens are not specified in the provided data.

Other GGUFs of MiniMax-H3

Similar models

Browse all →

More from webmp3

All by webmp3 →

More Minimax models

All Minimax →

Other medium models

All medium models →