LGAI-EXAONE_EXAONE-4.0-1.2B-GGUF Open-Source Language Model - Multi-Quantization Versions Compatible with Different Hardware

LGAI EXAONE EXAONE 4.0 1.2B GGUF

Developed by bartowski

EXAONE-4.0-1.2B is a 1.2B parameter language model released by LGAI-EXAONE, offering multiple quantization versions to meet different hardware requirements.

Large Language Model #Multiple quantization options #Low memory optimization #Online repackaging

Downloads 402

Release Time : 7/19/2025

Model Overview

This model is a 1.2B parameter language model quantized based on llama.cpp, providing multiple quantization types from BF16 to Q2_K, suitable for different hardware environments and performance requirements.

Model Features

Multiple quantization options

Provide multiple quantization types from BF16 to Q2_K to meet different hardware and performance requirements.

Embedded/output weight optimization

Some quantized models quantize the embedded and output weights to Q8_0 to improve model performance.

Online repackaging support

Some quantized models support online repackaging, which can improve performance on ARM and AVX machines.

Model Capabilities

Text generation

Dialogue system

Use Cases

Dialogue system

Intelligent assistant

Can be used to build intelligent dialogue assistants

Content generation

Text creation

Can be used to assist in text creation

🚀 Llamacpp imatrix Quantizations of EXAONE-4.0-1.2B by LGAI-EXAONE

This project provides quantized versions of the EXAONE-4.0-1.2B model using the llama.cpp library. These quantizations can significantly reduce the memory footprint of the model while maintaining reasonable performance, making it more accessible for various applications.

🚀 Quick Start

Using llama.cpp: Use llama.cpp release b5934 for quantization.
Running in LM Studio: You can run these quantized models in LM Studio.
Running directly with llama.cpp: Or you can run them directly with llama.cpp, or any other llama.cpp based project.

✨ Features

Multiple Quantization Types: Offers a wide range of quantization types, including bf16, Q8_0, Q6_K_L, etc., to meet different requirements for quality and memory usage.
Online Repacking: Some quantizations support online repacking for ARM and AVX machines, which can improve performance.
Customizable Download: Allows you to download specific files using the huggingface-cli.

📦 Installation

Downloading using huggingface-cli

Click to view download instructions

First, make sure you have hugginface-cli installed:

pip install -U "huggingface_hub[cli]"

Then, you can target the specific file you want:

huggingface-cli download bartowski/LGAI-EXAONE_EXAONE-4.0-1.2B-GGUF --include "LGAI-EXAONE_EXAONE-4.0-1.2B-Q4_K_M.gguf" --local-dir ./

If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download bartowski/LGAI-EXAONE_EXAONE-4.0-1.2B-GGUF --include "LGAI-EXAONE_EXAONE-4.0-1.2B-Q8_0/*" --local-dir ./

You can either specify a new local-dir (LGAI-EXAONE_EXAONE-4.0-1.2B-Q8_0) or download them all in place (./)

💻 Usage Examples

Prompt format

No chat template specified so default is used. This may be incorrect, check original model card for details.

[|system|]
{system_prompt}[|endofturn|]
[|user|]
{prompt}[|endofturn|]
[|assistant|]
<think>

</think>

📚 Documentation

Download a file (not the whole branch) from below:

Filename	Quant type	File Size	Split	Description
EXAONE-4.0-1.2B-bf16.gguf	bf16	2.56GB	false	Full BF16 weights.
EXAONE-4.0-1.2B-Q8_0.gguf	Q8_0	1.36GB	false	Extremely high quality, generally unneeded but max available quant.
EXAONE-4.0-1.2B-Q6_K_L.gguf	Q6_K_L	1.10GB	false	Uses Q8_0 for embed and output weights. Very high quality, near perfect, recommended.
EXAONE-4.0-1.2B-Q6_K.gguf	Q6_K	1.05GB	false	Very high quality, near perfect, recommended.
EXAONE-4.0-1.2B-Q5_K_L.gguf	Q5_K_L	0.98GB	false	Uses Q8_0 for embed and output weights. High quality, recommended.
EXAONE-4.0-1.2B-Q5_K_M.gguf	Q5_K_M	0.93GB	false	High quality, recommended.
EXAONE-4.0-1.2B-Q5_K_S.gguf	Q5_K_S	0.91GB	false	High quality, recommended.
EXAONE-4.0-1.2B-Q4_K_L.gguf	Q4_K_L	0.86GB	false	Uses Q8_0 for embed and output weights. Good quality, recommended.
EXAONE-4.0-1.2B-Q4_1.gguf	Q4_1	0.85GB	false	Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon.
EXAONE-4.0-1.2B-Q4_K_M.gguf	Q4_K_M	0.81GB	false	Good quality, default size for most use cases, recommended.
EXAONE-4.0-1.2B-Q3_K_XL.gguf	Q3_K_XL	0.79GB	false	Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability.
EXAONE-4.0-1.2B-Q4_K_S.gguf	Q4_K_S	0.78GB	false	Slightly lower quality with more space savings, recommended.
EXAONE-4.0-1.2B-Q4_0.gguf	Q4_0	0.78GB	false	Legacy format, offers online repacking for ARM and AVX CPU inference.
EXAONE-4.0-1.2B-IQ4_NL.gguf	IQ4_NL	0.78GB	false	Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference.
EXAONE-4.0-1.2B-IQ4_XS.gguf	IQ4_XS	0.75GB	false	Decent quality, smaller than Q4_K_S with similar performance, recommended.
EXAONE-4.0-1.2B-Q3_K_L.gguf	Q3_K_L	0.74GB	false	Lower quality but usable, good for low RAM availability.
EXAONE-4.0-1.2B-Q3_K_M.gguf	Q3_K_M	0.69GB	false	Low quality.
EXAONE-4.0-1.2B-IQ3_M.gguf	IQ3_M	0.66GB	false	Medium-low quality, new method with decent performance comparable to Q3_K_M.
EXAONE-4.0-1.2B-Q3_K_S.gguf	Q3_K_S	0.64GB	false	Low quality, not recommended.
EXAONE-4.0-1.2B-IQ3_XS.gguf	IQ3_XS	0.62GB	false	Lower quality, new method with decent performance, slightly better than Q3_K_S.
EXAONE-4.0-1.2B-Q2_K_L.gguf	Q2_K_L	0.62GB	false	Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable.
EXAONE-4.0-1.2B-Q2_K.gguf	Q2_K	0.57GB	false	Very low quality but surprisingly usable.
EXAONE-4.0-1.2B-IQ3_XXS.gguf	IQ3_XXS	0.56GB	false	Lower quality, new method with decent performance, comparable to Q3 quants.
EXAONE-4.0-1.2B-IQ2_M.gguf	IQ2_M	0.52GB	false	Relatively low quality, uses SOTA techniques to be surprisingly usable.

Embed/output weights

Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.

ARM/AVX information

Previously, you would download Q4_0_4_4/4_8/8_8, and these would have their weights interleaved in memory in order to improve performance on ARM and AVX machines by loading up more data in one pass.

Now, however, there is something called "online repacking" for weights. details in this PR. If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly.

As of llama.cpp build b4282 you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0.

Additionally, if you want to get slightly better quality for , you can use IQ4_NL thanks to this PR which will also repack the weights for ARM, though only the 4_4 for now. The loading time may be slower but it will result in an overall speed incrase.

Click to view Q4_0_X_X information (deprecated

I'm keeping this section to show the potential theoretical uplift in performance from using the Q4_0 with online repacking.

Click to view benchmarks on an AVX2 system (EPYC7702)

model	size	params	backend	threads	test	t/s	% (vs Q4_0)
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp512	204.03 ± 1.03	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp1024	282.92 ± 0.19	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp2048	259.49 ± 0.44	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg128	39.12 ± 0.27	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg256	39.31 ± 0.69	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg512	40.52 ± 0.03	100%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp512	301.02 ± 1.74	147%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp1024	287.23 ± 0.20	101%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp2048	262.77 ± 1.81	101%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg128	18.80 ± 0.99	48%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg256	24.46 ± 3.04	83%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg512	36.32 ± 3.59	90%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp512	271.71 ± 3.53	133%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp1024	279.86 ± 45.63	100%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp2048	320.77 ± 5.00	124%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg128	43.51 ± 0.05	111%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg256	43.35 ± 0.09	110%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg512	42.60 ± 0.31	105%

Q4_0_8_8 offers a nice bump to prompt processing and a small bump to text generation

Which file should I choose?

Click here for details

A great write up with charts showing various performances is provided by Artefact2 here

The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.

If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM.

If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total.

Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.

If you don't want to think too much, grab one of the K-quants. These are in format 'QX_K_X', like Q5_K_M.

If you want to get more into the weeds, you can check out this extremely useful feature chart:

llama.cpp feature matrix

But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQX_X, like IQ3_M. These are newer and offer better performance for their size.

These I-quants can also be used on CPU, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.