Instructions to use Vortex5/Starlit-Shadow-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Vortex5/Starlit-Shadow-12B with Transformers:

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

pipe = pipeline("text-generation", model="Vortex5/Starlit-Shadow-12B")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Vortex5/Starlit-Shadow-12B")
model = AutoModelForCausalLM.from_pretrained("Vortex5/Starlit-Shadow-12B")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Vortex5/Starlit-Shadow-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Vortex5/Starlit-Shadow-12B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Vortex5/Starlit-Shadow-12B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Vortex5/Starlit-Shadow-12B

SGLang

How to use Vortex5/Starlit-Shadow-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 "Vortex5/Starlit-Shadow-12B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Vortex5/Starlit-Shadow-12B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "Vortex5/Starlit-Shadow-12B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Vortex5/Starlit-Shadow-12B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Vortex5/Starlit-Shadow-12B with Docker Model Runner:
```
docker model run hf.co/Vortex5/Starlit-Shadow-12B
```

This is cool!

by MrDevolver - opened Dec 25, 2025

Discussion

MrDevolver

Dec 25, 2025

•

edited Dec 25, 2025

Hey Vortex5,

it's me again. When I saw Crimson Twilight in the mix, I wanted to give this a try and first response was almost perfect. I was using preset which worked well for similar scenarios with Crimson Twilight, so it looks like I won't need to tune the parameters too much, I just wanted to thank you for another cool model! I've tried many 12B models, but many times they are missing the ability to catch and interpret certain nuances properly. Your models of the same size got me covered with surprising capacity that reminds me of 24B Mistral Small finetunes!

I was wondering, have you ever done your own finetuning? If so, would you consider doing a Ministral 14B finetune? That model seems very capable, but so far I haven't seen many RP finetunes based on it. It feels like an upgrade for base Nemo, so I imagined that if you were able to make that 12B model feel like 24B, maybe you could do even better with 14B? 😉

Vortex5

Owner Dec 25, 2025

Thank you for the kind words. As for finetuning I was thinking of trying to do a very small model for the experience. As for the new ministral 14B I am waiting for someone to fine tune that 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