Instructions to use unsloth/GLM-5.2-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use unsloth/GLM-5.2-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="unsloth/GLM-5.2-GGUF",
	filename="BF16/GLM-5.2-BF16-00001-of-00033.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use unsloth/GLM-5.2-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 unsloth/GLM-5.2-GGUF:UD-Q4_K_M
# Run inference directly in the terminal:
llama cli -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M
# Run inference directly in the terminal:
llama cli -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Use Docker

docker model run hf.co/unsloth/GLM-5.2-GGUF:UD-Q4_K_M

LM Studio
Jan

vLLM

How to use unsloth/GLM-5.2-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "unsloth/GLM-5.2-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": "unsloth/GLM-5.2-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Ollama
How to use unsloth/GLM-5.2-GGUF with Ollama:
```
ollama run hf.co/unsloth/GLM-5.2-GGUF:UD-Q4_K_M
```

Unsloth Studio

How to use unsloth/GLM-5.2-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for unsloth/GLM-5.2-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for unsloth/GLM-5.2-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for unsloth/GLM-5.2-GGUF to start chatting

How to use unsloth/GLM-5.2-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "unsloth/GLM-5.2-GGUF:UD-Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use unsloth/GLM-5.2-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Run Hermes

hermes

Atomic Chat new
Docker Model Runner
How to use unsloth/GLM-5.2-GGUF with Docker Model Runner:
```
docker model run hf.co/unsloth/GLM-5.2-GGUF:UD-Q4_K_M
```

Lemonade

How to use unsloth/GLM-5.2-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull unsloth/GLM-5.2-GGUF:UD-Q4_K_M

Run and chat with the model

lemonade run user.GLM-5.2-GGUF-UD-Q4_K_M

List all available models

lemonade list

Issues with UD-Q2_K_XL quant

by labhraighlep - opened 4 days ago

Discussion

labhraighlep

4 days ago

•

edited 4 days ago

I was comparing some of these quants for quality and started with the lm-eval repo, the GSM8K bench specifically.
Ran to completion at 95% for the UD-IQ_4_NL and UD-Q4_K_S quants.
The UD-Q2_K_XL quant went into a runaway generation looping on decode at 37/1319 on 5 consecutive runs, always at the same spot. Reasoning was disabled. Ran the one shots for 36,37,38 separately, successfully but it failed out on the full test every time. Also tested limiting max_gen_toks.

Are there issues with this quant?

phakio

3 days ago

Just stopping by to say I noticed this as well, but I noticed it on a separate users Q2 quant... Some opencode sessions passed without a problem, even 100k+ context in, and others had this run on issue. It seemed to occur in the same spot. even after re-prompting from 0. Very annoying when it happens in a high context session.

I assume it was a run on generation loop because one of my 3 GPUS would peg at 100% and the others would idle. I have since grabbed a Q4 quant, and so far, although slower PP, haven't had this issue.

Chico70

about 23 hours ago

I have the same issue with UD-Q3_K_XL.... A prompt a bit more complex... And the model goes to an endless reasoning... Unfortunately, my setup doesn't handle higher quants.... :-(

phakio

about 11 hours ago

•

edited about 11 hours ago

@Chico70 - user @sokann has a great 2bit quant of GLM 5.2, and I got it to convert an entire .NET 3.5 codebase into a .NET 8.0 codebase. App now has cross compatibility and functions as the original version did. Might want to check that one out! I ran it over the course of 4 days, some sessions were 8+ hours and it worked GREAT.

labhraighlep

about 8 hours ago

•

edited about 8 hours ago

@chiko70
The GLM-5.2-GGUF/UD-IQ2_M version does not have this issue. I'm going to test the GLM-5.2-GGUF/UD-IQ1_M VERSION too to see if it is consistent. I'll test the UD-IQ3_S too because it falls under the size of the UD-Q3_K_KL you are seeing issues with. I'm doing this as part of my own effort to expand my abilities to be able to test quants.

I'm using the GSM8K benchmarks testing the looping running through all 3139 tests,

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