Instructions to use Mungert/EXAONE-Deep-32B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Mungert/EXAONE-Deep-32B-GGUF with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Mungert/EXAONE-Deep-32B-GGUF")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Mungert/EXAONE-Deep-32B-GGUF", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use Mungert/EXAONE-Deep-32B-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use Mungert/EXAONE-Deep-32B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Mungert/EXAONE-Deep-32B-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Mungert/EXAONE-Deep-32B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M

SGLang

How to use Mungert/EXAONE-Deep-32B-GGUF with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Mungert/EXAONE-Deep-32B-GGUF" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Mungert/EXAONE-Deep-32B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Mungert/EXAONE-Deep-32B-GGUF" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Mungert/EXAONE-Deep-32B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use Mungert/EXAONE-Deep-32B-GGUF with Ollama:
```
ollama run hf.co/Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M
```
Unsloth Desktop
Docker Model Runner
How to use Mungert/EXAONE-Deep-32B-GGUF with Docker Model Runner:
```
docker model run hf.co/Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M
```

Lemonade

How to use Mungert/EXAONE-Deep-32B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Mungert/EXAONE-Deep-32B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.EXAONE-Deep-32B-GGUF-Q4_K_M

List all available models

lemonade list

Atomic Chat

Updated Model

by D0schie - opened Jun 25, 2025

Discussion

D0schie

Jun 25, 2025

At first glance the IQ1_S model stopped to loop indefinitely after i asked it about itself so thanks for the update!

Mungert

Owner Jun 25, 2025

At first glance the IQ1_S model stopped to loop indefinitely after i asked it about itself so thanks for the update!
Thanks for feedback

The iq1_s quant has not updated yet. Unfortunately my hardware is limited so its slow. Check on the file update time in about 10 hours maybe. I would be interesting to see if the test you made with the iq1_s model shows improvements with the new 1q_s quant.

D0schie

Jun 25, 2025

Oh i see dont worry i will give feedback! Do you btw know if i can just update the LLM in LMStudio or do i need to download the whole thing again?

D0schie

Jun 25, 2025

are you doing all of this work just on a zen4?!

Mungert

Owner Jun 25, 2025

•

edited Jun 25, 2025

are you doing all of this work just on a zen4?!

Yeah, well not exactly i use two vps hired from contabo.com . One VPS with 6GB ram 4 threads of a AMD EPYC 9224 this has avx512, other one 12GB ram 6 threads of a Intel Core Processor Broadwell only has avx2. They run about the same speed . Adds up to less compute than a Zen4 but there is no lower selection in the profile page. That's all I can afford at the moment :(

Mungert

Owner Jun 25, 2025

Oh i see dont worry i will give feedback! Do you btw know if i can just update the LLM in LMStudio or do i need to download the whole thing again?

You will need to download the whole gguf file again as it will have different weights for pretty much all the tensors.

D0schie

Jun 25, 2025

are you doing all of this work just on a zen4?!

Yeah, well not exactly i use two vps hired from contabo.com . One VPS with 6GB ram 4 threads of a AMD EPYC 9224 this has avx512, other one 12GB ram 6 threads of a Intel Core Processor Broadwell only has avx2. They run about the same speed . Adds up to less compute than a Zen4 but there is no lower selection in the profile page. That's all I can afford at the moment :(

nah thats awesome keep going!

Ricmod

Jun 29, 2025

are you doing all of this work just on a zen4?!

Yeah, well not exactly i use two vps hired from contabo.com . One VPS with 6GB ram 4 threads of a AMD EPYC 9224 this has avx512, other one 12GB ram 6 threads of a Intel Core Processor Broadwell only has avx2. They run about the same speed . Adds up to less compute than a Zen4 but there is no lower selection in the profile page. That's all I can afford at the moment :(

Make sure you're using AVX512 builds on the epyc. avx512 support can put cpu performance within an order of magnitude or two of a gpu.

Mungert

Owner Jun 29, 2025

are you doing all of this work just on a zen4?!

Yeah, well not exactly i use two vps hired from contabo.com . One VPS with 6GB ram 4 threads of a AMD EPYC 9224 this has avx512, other one 12GB ram 6 threads of a Intel Core Processor Broadwell only has avx2. They run about the same speed . Adds up to less compute than a Zen4 but there is no lower selection in the profile page. That's all I can afford at the moment :(

Make sure you're using AVX512 builds on the epyc. avx512 support can put cpu performance within an order of magnitude or two of a gpu.

Yes building openblas on the epyc hosts machine detects and builds with avx512. That helps with prompt processing. Not totally sure but I think llama.cpp also builds with avx512 as well

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment