Instructions to use Situus/STARK-WEB-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Situus/STARK-WEB-12B with Transformers:

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

pipe = pipeline("text-generation", model="Situus/STARK-WEB-12B")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("Situus/STARK-WEB-12B")
model = AutoModelForMultimodalLM.from_pretrained("Situus/STARK-WEB-12B")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Situus/STARK-WEB-12B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Situus/STARK-WEB-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": "Situus/STARK-WEB-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Situus/STARK-WEB-12B

SGLang

How to use Situus/STARK-WEB-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 "Situus/STARK-WEB-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": "Situus/STARK-WEB-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 "Situus/STARK-WEB-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": "Situus/STARK-WEB-12B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Situus/STARK-WEB-12B with Docker Model Runner:
```
docker model run hf.co/Situus/STARK-WEB-12B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

STARK-WEB-12B (v1.0 Legacy)

💡 NOTE: This is the initial version 1.0 of STARK-WEB-12B. While it laid the foundation for our project, we highly recommend checking out the newest v1.7 on our main repository, which features massive improvements to reasoning, dataset quality, and game development.

Overview

STARK-WEB-12B v1.0 was our first step in fine-tuning the Gemma 4 12B architecture to act as a frontend web developer and UI designer.

It is capable of generating structured HTML, CSS, and basic JavaScript in a single file, and can successfully build simple websites, dashboards, and forms.

Limitations of this Version

As an early experimental release, this model has a few known quirks that users should be aware of:

Overfitting Tendencies: Because the initial dataset was smaller and less refined, the model sometimes memorizes specific code blocks. If you prompt it for very niche applications, it might fall back to patterns it knows too well instead of adapting creatively.
Reasoning (CoT) Overhead: The internal reasoning process is present, but occasionally the model might overthink simple problems or get stuck in repetitive logic loops.
Complex Logic & Games: While it handles UI well, creating complex interactive games (like Canvas mechanics or physics) can result in bugs or missing event handlers.

Specifications

Base Architecture: Gemma 4 12B
Status: Legacy / v1.0
Format: GGUF (Ready for local execution)

Conclusion

This version is fully usable for basic web tasks and serves as a great historical milestone for the STARK-WEB project. However, for serious prototyping and bug-free logic, we strongly encourage upgrading to v1.7.

Downloads last month: 60

Safetensors

Model size

12B params

Tensor type

BF16

Model tree for Situus/STARK-WEB-12B

Quantizations

2 models

Collection including Situus/STARK-WEB-12B

STARK-WEB

Collection

4 items • Updated 6 days ago