Instructions to use WebUniverse-NSL/Qwen3.5-StepFun-Mix with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use WebUniverse-NSL/Qwen3.5-StepFun-Mix with Transformers:

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

pipe = pipeline("image-text-to-text", model="WebUniverse-NSL/Qwen3.5-StepFun-Mix")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("WebUniverse-NSL/Qwen3.5-StepFun-Mix")
model = AutoModelForMultimodalLM.from_pretrained("WebUniverse-NSL/Qwen3.5-StepFun-Mix", 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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use WebUniverse-NSL/Qwen3.5-StepFun-Mix with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "WebUniverse-NSL/Qwen3.5-StepFun-Mix"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "WebUniverse-NSL/Qwen3.5-StepFun-Mix",
		"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/WebUniverse-NSL/Qwen3.5-StepFun-Mix

SGLang

How to use WebUniverse-NSL/Qwen3.5-StepFun-Mix 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 "WebUniverse-NSL/Qwen3.5-StepFun-Mix" \
    --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": "WebUniverse-NSL/Qwen3.5-StepFun-Mix",
		"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 "WebUniverse-NSL/Qwen3.5-StepFun-Mix" \
        --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": "WebUniverse-NSL/Qwen3.5-StepFun-Mix",
		"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 WebUniverse-NSL/Qwen3.5-StepFun-Mix with Docker Model Runner:
```
docker model run hf.co/WebUniverse-NSL/Qwen3.5-StepFun-Mix
```

Access Qwen3.5-StepFun-Mix

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

By requesting access, you agree to the Qwen3.5-StepFun-Mix License, including noncommercial use of our fine-tuning contributions, and acknowledge the StepFun data terms linked in the model card. The model is provided as is, without warranty.

Qwen3.5-StepFun-Mix

Public-data comparison for WebUniverse.

Qwen3.5-StepFun-Mix is a fine-tuned Qwen3.5-35B-A3B model for agentic web development.

Training data

We trained this comparison model on StepFun-Mix, our 15,000-example subset of StepFun's Step-3.5-Flash-SFT, selected from json/general/:

Component	Examples
Front-end webpage generation	13,500
Multi-turn software engineering	1,500

We filtered and deduplicated the examples, screened for evaluation overlap, and converted them to our training format. The model uses the same base checkpoint and training configuration as Qwen3.5-WebUniverse. StepFun-Mix names our selected subset; the source corpus was released by StepFun.

Training configuration

Setting	Value
Base model	Qwen3.5-35B-A3B
Framework	Megatron-LM through ms-swift
Epochs	2
Global batch size	16
Learning rate	2e-5
Schedule	Cosine decay, 10% warmup
Packed sequence length	98,304 tokens
Input length limit	106,496 tokens
Precision	BF16
Vision tower	Frozen

Evaluation results

Model	Vision2Web L1 average	WebGen-Bench overall (%)	Web-Bench pass@1 (%)	Web-Bench pass@2 (%)
Qwen3.5-35B-A3B	51.7	34.5	7.2	12.2
Qwen3.5-StepFun-Mix	52.4	31.5	9.3	17.7

Evaluated with OpenCode benchmark adapters and vLLM 0.24.0, using temperature 0.6, top-p 0.95, and top-k 20.

Evaluation code and full configurations will be released with the project code.

Quick start

Tested with vLLM 0.24.0:

vllm serve WebUniverse-NSL/Qwen3.5-StepFun-Mix \
  --host 127.0.0.1 \
  --port 8000 \
  --served-model-name Qwen3.5-StepFun-Mix \
  --tensor-parallel-size 8 \
  --max-model-len 98304 \
  --reasoning-parser qwen3 \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_coder

Adjust tensor parallelism and context length to your GPU memory. See the base model documentation for serving details.

Resources

Resource	Location
Training data source	Step-3.5-Flash-SFT
WebUniverse comparison dataset	WebUniverse-Mix
Models	WebUniverse Models
Task environments	WebUniverse-Env: executable tasks, initial workspaces, and verifiers