MI

MiniMax H3 Hybrid b25 49

MinimaxGGUF by hoidhxdUpdated 2026-08-13
Download on Hugging Face
Parameters20.1B
Context—
Layers—
Architectureminimax_h3
Smallest quant11.4 GB
Quick answer

MiniMax H3 Hybrid b25 49 needs about 13.3 GB of memory at Q4_0 (a 11.4 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 2 quants, from 11.4 GB (Q4_0) to 13.9 GB (Q5_0).

MiniMax H3 Hybrid b25 49 – a merged audio+video diffusion model

MiniMax H3 Hybrid b25 49 merges the FL2VA and REF2VA checkpoints. It keeps FL2VA’s high visual and audio quality while adding REF2VA reference conditioning. The model is 20.1 billion parameters and is available in Q4_0 and Q5_0 GGUF formats.

Good at

  • High‑quality audio‑video generation
  • Reference‑driven conditioning
  • Balanced visual fidelity

Limits

  • No vision support beyond audio‑video
  • Large VRAM requirement for Q5_0
  • No license information provided

Best for

  • Audio‑video synthesis
  • Reference‑guided content creation
  • Multimodal generation tasks

Who it’s for. Users needing a single model that supports both high‑quality audio‑video output and reference conditioning. Suitable for developers building multimodal generation pipelines.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_011.4 GB 13.3 GB
Too big Get file →
Q5_013.9 GB 16.1 GB
Too big Get file →

For 8 GB VRAM, use the Q4_0 file (11.39 GB, ~13.3 GB VRAM at 8k context). If higher precision is needed and you have 12 GB VRAM, the Q5_0 file (13.9 GB, ~16.1 GB VRAM at 8k context) is recommended. Estimates are approximate.

What GPU do you need for MiniMax H3 Hybrid b25 49?

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_0 (11.4 GB file, about 13.3 GB in use).
  • 24 GB VRAM: the best quant that fits is Q5_0 (13.9 GB file, about 16.1 GB in use).

Which MiniMax H3 Hybrid b25 49 quant should you choose?

Q4_0 (11.4 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureminimax_h3
Parameters20.1B

Source

Quantized byhoidhxd
FamilyMinimax
Downloads (HF)1,947
Last updated2026-08-13

Frequently asked questions about MiniMax H3 Hybrid b25 49

How much VRAM does MiniMax H3 Hybrid b25 49 need?

About 13.3 GB at Q4_0 with an 8K context window. That is the 11.4 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 MiniMax H3 Hybrid b25 49 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 13.3 GB, so you would need a larger GPU or CPU offloading.

What is the file size of MiniMax H3 Hybrid b25 49 GGUF?

It depends on the quant: Q4_0 is 11.4 GB, Q5_0 is 13.9 GB.

Where can I download MiniMax H3 Hybrid b25 49 GGUF?

The files are on Hugging Face in the hoidhxd/MiniMax-H3-Hybrid-b25-49-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Can I run MiniMax H3 Hybrid on 8 GB VRAM?

Yes, with the Q4_0 quantization which requires about 13.3 GB VRAM at 8k context. It may need to run with a smaller context or batch size.

Does this model support text‑only generation?

The model is designed for audio‑video diffusion and reference conditioning; text‑only generation is not its primary use case.

What quantization gives better quality?

Q5_0 offers slightly better precision than Q4_0, but it requires more VRAM.

Similar models

Browse all →

More from hoidhxd

All by hoidhxd →

More Minimax models

All Minimax →

Other medium models

All medium models →