LL

Llama 3.2 3B MoE 4Expert

LlamaGGUF by mradermacherUpdated 2026-02-18
Download on Hugging Face
Parameters9.55B
Context128K
Layers28
Architecturellama
Smallest quant0.0 GB
Quick answer

Llama 3.2 3B MoE 4Expert needs about 7.3 GB of memory at Q4_K_M (a 5.9 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 25 quants, from 0.0 GB to 7.8 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 3.8 GB.

Llama 3.2 3B MoE 4Expert – 4‑expert mixture for chat, code, creative writing and math

Llama 3.2 3B MoE 4Expert is a 4‑expert mixture built on unsloth/Llama‑3.2‑3B‑Instruct. It routes queries to specialized experts for general chat, code, creative writing, and mathematics. The model supports a 131072‑token context window.

Good at

  • Expert routing for chat, code, creative writing, math
  • Large context of 131072 tokens
  • Mixture of experts architecture

Limits

  • 3B parameter size limits depth
  • Requires prompt phrasing for correct routing
  • No vision support

Best for

  • Chatting with structured answers
  • Programming assistance
  • Creative writing prompts

Who it’s for. Developers and writers who need a single model that can handle code, creative content, and math. Useful for applications that benefit from expert routing without switching models.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.0 GB 1.4 GB
Fits Get file →
IQ1_S2.2 GB 3.6 GB
Fits Get file →
IQ1_M2.4 GB 3.8 GB
Fits Get file →
IQ2_XXS2.7 GB 4.1 GB
Fits Get file →
IQ2_XS3.0 GB 4.4 GB
Fits Get file →
IQ2_S3.0 GB 4.5 GB
Fits Get file →
IQ2_M3.3 GB 4.7 GB
Fits Get file →
Q2_K_S3.4 GB 4.8 GB
Fits Get file →
Q2_K3.7 GB 5.1 GB
Fits Get file →
IQ3_XXS3.8 GB 5.2 GB
Fits Get file →
IQ3_XS4.1 GB 5.5 GB
Fits Get file →
IQ3_S4.3 GB 5.7 GB
Fits Get file →
Q3_K_S4.3 GB 5.7 GB
Fits Get file →
IQ3_M4.4 GB 5.8 GB
Fits Get file →
Q3_K_M4.7 GB 6.1 GB
Fits Get file →
Q3_K_L5.1 GB 6.5 GB
Fits Get file →
IQ4_XS5.2 GB 6.6 GB
Fits Get file →
IQ4_NL5.5 GB 6.9 GB
Fits Get file →
Q4_05.5 GB 6.9 GB
Fits Get file →
Q4_K_S5.5 GB 7.0 GB
Fits Get file →
Q4_K_M5.9 GB 7.3 GB
Tight Get file →
Q4_16.1 GB 7.5 GB
Tight Get file →
Q5_K_S6.6 GB 8.1 GB
Too big Get file →
Q5_K_M6.8 GB 8.3 GB
Too big Get file →
Q6_K7.8 GB 9.3 GB
Too big Get file →

For 8 GB VRAM use IQ1_S (2.16 GB, 3.6 GB memory) or IQ1_M (2.36 GB, 3.8 GB memory). With 12 GB VRAM, Q2_K (3.66 GB, 5.1 GB memory) or Q3_K_S (4.27 GB, 5.7 GB memory) are suitable. Larger VRAM allows higher‑quality quants such as Q5_K_M (6.82 GB, 8.3 GB memory).

What GPU do you need for Llama 3.2 3B MoE 4Expert?

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: the best quant that fits is IQ1_M (2.4 GB file, about 3.8 GB in use).
  • 6 GB VRAM: the best quant that fits is IQ3_M (4.4 GB file, about 5.8 GB in use).
  • 8 GB VRAM: the best quant that fits is Q4_1 (6.1 GB file, about 7.5 GB in use).
  • 12 GB VRAM: the best quant that fits is Q6_K (7.8 GB file, about 9.3 GB in use).

Which Llama 3.2 3B MoE 4Expert quant should you choose?

Q4_K_M is the usual starting point for Llama 3.2 3B MoE 4Expert: it balances file size and quality well. Here it is a 5.9 GB file.

If memory is tight, the smallest file is the first one in the table at 0.0 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

Architecturellama
Parameters9.55B (approximate, mixture of experts)
Layers28
Trained context131,072 tokens
KV cache (fp16)112 KB per token

Source

Quantized bymradermacher
FamilyLlama
Licensellama3.2
Downloads (HF)2,706
Last updated2026-02-18

Frequently asked questions about Llama 3.2 3B MoE 4Expert

How much VRAM does Llama 3.2 3B MoE 4Expert need?

About 7.3 GB at Q4_K_M with an 8K context window. That is the 5.9 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 Llama 3.2 3B MoE 4Expert on an 8 GB GPU?

Yes, with Q4_1: the 6.1 GB file needs about 7.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Llama 3.2 3B MoE 4Expert GGUF?

It depends on the quant: file is 0.0 GB, IQ1_S is 2.2 GB, IQ1_M is 2.4 GB, IQ2_XXS is 2.7 GB, IQ2_XS is 3.0 GB, IQ2_S is 3.0 GB and 19 more.

What context length does Llama 3.2 3B MoE 4Expert support?

The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.9 GB at 8K and 3.8 GB at 32K.

Where can I download Llama 3.2 3B MoE 4Expert GGUF?

The files are on Hugging Face in the mradermacher/Llama-3.2-3B-MoE-4Expert-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 Llama 3.2 3B MoE 4Expert use?

Hugging Face lists the license as llama3.2. 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 this model on 8 GB VRAM?

Yes, use IQ1_S or IQ1_M quant files; they require about 3.6–3.8 GB VRAM at 8k context.

Does the model support vision?

No, the model has no vision capabilities.

What context length does it support?

It supports up to 131072 tokens in context.

Other GGUFs of Llama-3.2-3B-MoE-4Expert

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Llama models

All Llama →

Other medium models

All medium models →