Instructions to use microsoft/GELab-Zero-4B-preview-Sico-Evolution with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use microsoft/GELab-Zero-4B-preview-Sico-Evolution with Transformers:

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

pipe = pipeline("image-text-to-text", model="microsoft/GELab-Zero-4B-preview-Sico-Evolution")
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("microsoft/GELab-Zero-4B-preview-Sico-Evolution")
model = AutoModelForMultimodalLM.from_pretrained("microsoft/GELab-Zero-4B-preview-Sico-Evolution", device_map="auto")
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 microsoft/GELab-Zero-4B-preview-Sico-Evolution with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "microsoft/GELab-Zero-4B-preview-Sico-Evolution"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "microsoft/GELab-Zero-4B-preview-Sico-Evolution",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/microsoft/GELab-Zero-4B-preview-Sico-Evolution

SGLang

How to use microsoft/GELab-Zero-4B-preview-Sico-Evolution 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 "microsoft/GELab-Zero-4B-preview-Sico-Evolution" \
    --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": "microsoft/GELab-Zero-4B-preview-Sico-Evolution",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "microsoft/GELab-Zero-4B-preview-Sico-Evolution" \
        --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": "microsoft/GELab-Zero-4B-preview-Sico-Evolution",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use microsoft/GELab-Zero-4B-preview-Sico-Evolution with Docker Model Runner:
```
docker model run hf.co/microsoft/GELab-Zero-4B-preview-Sico-Evolution
```

GELab-Zero-4B-preview-Sico-Evolution / README.md

Experiencency

Tighten base model address spacing

c92fa07 verified 24 days ago

preview code

Raw

History Blame Contribute Delete

1.88 kB

	---
	license: apache-2.0
	base_model: GELab-Zero-4B-preview
	library_name: transformers
	pipeline_tag: image-text-to-text
	language:
	- en
	- zh
	tags:
	- gui-agent
	- mobile-agent
	- vision-language
	- qwen3-vl
	- lora
	---

	> - 🔧 This model is part of [Sico](https://github.com/microsoft/Sico) — an open-source platform for building and evolving Digital Workers, where AI agents and their human operators co-evolve through real work.
	> - ⭐ [Star the Sico repository](https://github.com/microsoft/Sico) to follow new evolved models and our model evolution pipeline — this GUI agent is the first public release, with more on the way.
	> - 📄 Backed by our survey on [agentic evolution and co-evolving human–AI systems](https://www.microsoft.com/en-us/research/publication/agentic-evolution-from-self-improving-agents-to-co-evolving-human-ai-systems/).

	# GELab-Zero-4B-preview-Sico-Evolution

	A 4B GUI agent fine-tuned (LoRA) from the open-source GELab-Zero-4B-preview base model
	on Microsoft Edge and Copilot UI trajectories. It is built with our **general-purpose
	GUI model evolution pipeline** — an iterative mechanism that keeps lifting an agent's real
	task success rate round after round, and transfers to any GUI app.
	Base model address: https://huggingface.co/stepfun-ai/GELab-Zero-4B-preview

	## Highlights

	From 39.8% to 82.9%: Sico-Evolution achieves a dominant **82.9%
	Task Success Rate, a massive +43.1% absolute surge over the 39.8%** base-model
	baseline.

	Outperforms Closed-Source SOTAs: It edges out top proprietary giants like **gpt-5.4
	(79.7%), Claude-Opus-4.6 (81.3%), and claude-opus-4.7 (82.1%)**.

	Vastly Exceeds Open-Source Models: It crushes leading competitors including
	kimi-k2.6 (62.6%) and UI-Venus-1.5-30B (61.0%).

	## Results

	![Edge / Copilot Test Cases — TSR](tsr_edge_copilot.png)