Instructions to use KaraKaraWitch/Llama-3.3-MagicalGirl-2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use KaraKaraWitch/Llama-3.3-MagicalGirl-2 with Transformers:

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

pipe = pipeline("text-generation", model="KaraKaraWitch/Llama-3.3-MagicalGirl-2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("KaraKaraWitch/Llama-3.3-MagicalGirl-2")
model = AutoModelForCausalLM.from_pretrained("KaraKaraWitch/Llama-3.3-MagicalGirl-2", device_map="auto")
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 Settings

vLLM

How to use KaraKaraWitch/Llama-3.3-MagicalGirl-2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "KaraKaraWitch/Llama-3.3-MagicalGirl-2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "KaraKaraWitch/Llama-3.3-MagicalGirl-2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/KaraKaraWitch/Llama-3.3-MagicalGirl-2

SGLang

How to use KaraKaraWitch/Llama-3.3-MagicalGirl-2 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 "KaraKaraWitch/Llama-3.3-MagicalGirl-2" \
    --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": "KaraKaraWitch/Llama-3.3-MagicalGirl-2",
		"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 "KaraKaraWitch/Llama-3.3-MagicalGirl-2" \
        --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": "KaraKaraWitch/Llama-3.3-MagicalGirl-2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use KaraKaraWitch/Llama-3.3-MagicalGirl-2 with Docker Model Runner:
```
docker model run hf.co/KaraKaraWitch/Llama-3.3-MagicalGirl-2
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

New merge. This an experiment to increase the "Madness" in a model. Merge is based on top UGI-Bench models (So yeah, I would think this would be benchmaxxing.)

This is the second time I'm using SCE. The previous MagicalGirl model seems to be quite happy with it.

Added KaraKaraWitch/Llama-MiraiFanfare-3.3-70B based on feedback I got from others (People generally seem to remember this rather than other models). So I'm not sure how this would play into the merge.

This is a merge of pre-trained language models created using mergekit.

UGI-Results

Pretty interesting. As of 05/03/25, it's in the top 10th:

Bench	Results
UGI-Score	52.48 / 100
Unruly	3.8 / 10
Internet	5.1 / 10
Society	5.4 / 10
Willing	7 / 10
NatInt	41.86 / 100
Coding	22
Politial Lean	−3.9% (Liberalism)

Merge Details

Merge Method

This model was merged using the SCE merge method using KaraKaraWitch/Llama-3.X-Workout-70B as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: SicariusSicariiStuff/Negative_LLAMA_70B
  - model: TheDrummer/Anubis-70B-v1
  - model: KaraKaraWitch/Llama-MiraiFanfare-3.3-70B
  - model: Black-Ink-Guild/Pernicious_Prophecy_70B
  - model: LatitudeGames/Wayfarer-Large-70B-Llama-3.3

merge_method: sce
base_model: KaraKaraWitch/Llama-3.X-Workout-70B
parameters:
  select_topk: 1.0
dtype: bfloat16