Instructions to use PygmalionAI/Pygmalion-3-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use PygmalionAI/Pygmalion-3-12B with Transformers:

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

pipe = pipeline("text-generation", model="PygmalionAI/Pygmalion-3-12B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("PygmalionAI/Pygmalion-3-12B")
model = AutoModelForCausalLM.from_pretrained("PygmalionAI/Pygmalion-3-12B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use PygmalionAI/Pygmalion-3-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "PygmalionAI/Pygmalion-3-12B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "PygmalionAI/Pygmalion-3-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/PygmalionAI/Pygmalion-3-12B

SGLang

How to use PygmalionAI/Pygmalion-3-12B 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 "PygmalionAI/Pygmalion-3-12B" \
    --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": "PygmalionAI/Pygmalion-3-12B",
		"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 "PygmalionAI/Pygmalion-3-12B" \
        --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": "PygmalionAI/Pygmalion-3-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use PygmalionAI/Pygmalion-3-12B with Docker Model Runner:
```
docker model run hf.co/PygmalionAI/Pygmalion-3-12B
```

This model is amazing

by zenoran - opened Apr 8, 2025

Discussion

zenoran

Apr 8, 2025

•

edited Apr 8, 2025

I just wanted a model that could tell stories without limits. I've tried other LLMs and they are extremely difficult to get unique and creative responses. Not only can this get off with a great start from practically nothing, but it is also excellent at taking guidance as it iterates through endless story and dialog. This thing is writing novels, taking simple guidance and flowing freely with intriguing and coherent stories. I have this paired up with the CSM 1b TTS model and ... Wow...

The other amazing thing is it won't fight with you on what you want it to write about. It doesn't give BS disclaimers and try to gaslight you into thinking you're going to tear down the fabric of humanity with your immoral requests for entertainment. I haven't tried previous versions of this model but I'm yet to run into anything anywhere near the freedom and flexibility this provides.

All that being said, performance is a bit confusing for me. Running on a 4090 I've only been able to get it to push around 40% clock and it's very slow to spit out tokens. Would be interested to know if there's some trick to get this thing to kick it up a notch.

All in all... very happy finding this model. Looking forward to playing around with it more and seeing what all it can do.

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