HU

Huihui4 48B A4B Abliterated

HuihuiGGUF by mradermacherUpdated 2026-04-20
Download on Hugging Face
Parameters48B
Context256K
Layers30
Architecturegemma4
Smallest quant0.1 GB
Quick answer

Huihui4 48B A4B Abliterated needs about 40.5 GB of memory at Q4_K_M (a 32.0 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 24 quants, from 0.1 GB to 43.3 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 32.2 GB.

Huihui4 48B A4B Abliterated – a 48B Mixture of Experts Gemma4 model

Huihui4 48B A4B Abliterated is a 48‑billion‑parameter Mixture of Experts language model built on Gemma4. It activates 4 billion parameters per inference and supports 262 144‑token context. The model is released under apache‑2.0.

Good at

  • Large‑context text generation
  • Question answering
  • Conversational AI

Limits

  • Requires 8 k context for memory estimates
  • No vision support in this GGUF release
  • Fine‑tuning recommended for best performance

Best for

  • Research on MoE architectures
  • High‑volume text generation
  • Chatbot back‑end

Who it’s for. Researchers and developers needing a large‑context MoE model. The model is suitable for those who can fine‑tune and manage the memory requirements.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.1 GB 8.7 GB
Too big Get file →
IQ1_S15.6 GB 24.1 GB
Too big Get file →
IQ1_M16.3 GB 24.9 GB
Too big Get file →
IQ2_XXS17.5 GB 26.1 GB
Too big Get file →
IQ2_XS18.5 GB 27.1 GB
Too big Get file →
IQ2_S18.6 GB 27.2 GB
Too big Get file →
IQ2_M19.6 GB 28.1 GB
Too big Get file →
Q2_K19.9 GB 28.5 GB
Too big Get file →
Q2_K_S20.0 GB 28.6 GB
Too big Get file →
IQ3_XXS21.5 GB 30.0 GB
Too big Get file →
IQ3_XS22.0 GB 30.5 GB
Too big Get file →
IQ3_S23.1 GB 31.6 GB
Too big Get file →
Q3_K_S23.1 GB 31.6 GB
Too big Get file →
IQ3_M23.4 GB 31.9 GB
Too big Get file →
Q3_K_M25.1 GB 33.7 GB
Too big Get file →
Q3_K_L26.1 GB 34.7 GB
Too big Get file →
IQ4_XS26.3 GB 34.9 GB
Too big Get file →
Q4_027.4 GB 36.0 GB
Too big Get file →
Q4_K_S29.4 GB 37.9 GB
Too big Get file →
Q4_130.3 GB 38.8 GB
Too big Get file →
Q4_K_M32.0 GB 40.5 GB
Too big Get file →
Q5_K_S34.2 GB 42.8 GB
Too big Get file →
Q5_K_M36.5 GB 45.0 GB
Too big Get file →
Q6_K43.3 GB 51.8 GB
Too big Get file →

For 8 GB VRAM use IQ1_S (15.59 GB, 24.1 GB memory). For 16 GB VRAM use Q2_K (19.9 GB, 28.5 GB memory). For 32 GB VRAM use Q5_K_M (36.46 GB, 45 GB memory). Estimates are approximate.

What GPU do you need for Huihui4 48B A4B Abliterated?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) 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: the best quant that fits is the main file (0.1 GB file, about 8.7 GB in use).
  • 16 GB VRAM: the best quant that fits is the main file (0.1 GB file, about 8.7 GB in use).
  • 24 GB VRAM: the best quant that fits is the main file (0.1 GB file, about 8.7 GB in use).
  • 32 GB VRAM: the best quant that fits is IQ3_M (23.4 GB file, about 31.9 GB in use).
  • 48 GB VRAM: the best quant that fits is Q5_K_M (36.5 GB file, about 45.0 GB in use).

Which Huihui4 48B A4B Abliterated quant should you choose?

Q4_K_M is the usual starting point for Huihui4 48B A4B Abliterated: it balances file size and quality well. Here it is a 32.0 GB file.

If memory is tight, the smallest file is the first one in the table at 0.1 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturegemma4
Parameters48B (approximate, mixture of experts)
Active parameters4B
Layers30
Trained context262,144 tokens
KV cache (fp16)960 KB per token

Source

Quantized bymradermacher
FamilyHuihui
Licenseapache-2.0
Downloads (HF)1,809
Last updated2026-04-20

Frequently asked questions about Huihui4 48B A4B Abliterated

How much VRAM does Huihui4 48B A4B Abliterated need?

About 40.5 GB at Q4_K_M with an 8K context window. That is the 32.0 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 Huihui4 48B A4B Abliterated on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 8.7 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Huihui4 48B A4B Abliterated GGUF?

It depends on the quant: file is 0.1 GB, IQ1_S is 15.6 GB, IQ1_M is 16.3 GB, IQ2_XXS is 17.5 GB, IQ2_XS is 18.5 GB, IQ2_S is 18.6 GB and 18 more.

What context length does Huihui4 48B A4B Abliterated support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 8.1 GB at 8K and 32.2 GB at 32K.

Where can I download Huihui4 48B A4B Abliterated GGUF?

The files are on Hugging Face in the mradermacher/Huihui4-48B-A4B-abliterated-i1-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 Huihui4 48B A4B Abliterated use?

Hugging Face lists the license as apache-2.0. 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 Huihui4 48B on 8 GB VRAM?

Use the IQ1_S quant; it requires about 24.1 GB memory at 8 k context, so 8 GB VRAM is insufficient.

Does this model support image inputs?

No, the GGUF files provided are text‑only; vision support is not included.

Is fine‑tuning required?

Fine‑tuning is recommended to optimize expert routing, but the model can run out‑of‑the‑box for general tasks.

Other GGUFs of Huihui4-48B-A4B-abliterated

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Huihui models

All Huihui →

Other large models

All large models →