Instructions to use Qwen/WebWorld-32B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Qwen/WebWorld-32B with Transformers:

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

pipe = pipeline("text-generation", model="Qwen/WebWorld-32B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Qwen/WebWorld-32B")
model = AutoModelForCausalLM.from_pretrained("Qwen/WebWorld-32B")
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]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Qwen/WebWorld-32B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Qwen/WebWorld-32B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Qwen/WebWorld-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Qwen/WebWorld-32B

SGLang

How to use Qwen/WebWorld-32B 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 "Qwen/WebWorld-32B" \
    --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": "Qwen/WebWorld-32B",
		"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 "Qwen/WebWorld-32B" \
        --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": "Qwen/WebWorld-32B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Qwen/WebWorld-32B with Docker Model Runner:
```
docker model run hf.co/Qwen/WebWorld-32B
```

VoyagerXHF commited on 8 days ago

Commit

e7fceed

verified ·

1 Parent(s): d294637

Update README.md

Browse files

Files changed (1) hide show

README.md +17 -7

README.md CHANGED Viewed

@@ -29,7 +29,17 @@ datasets:
 # WebWorld 🌐
-[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/LICENSE-2.0) [![GitHub](https://img.shields.io/badge/GitHub-WebWorld-4b32c3?logo=github)](https://github.com/QwenLM/WebWorld) [![Dataset](https://img.shields.io/badge/HF%20Dataset-WebWorldData-yellow?logo=huggingface)](https://huggingface.co/datasets/Qwen/WebWorldData) [![8B](https://img.shields.io/badge/Model-8B-green?logo=huggingface)](https://huggingface.co/Qwen/WebWorld-8B) [![14B](https://img.shields.io/badge/Model-14B-green?logo=huggingface)](https://huggingface.co/Qwen/WebWorld-14B) [![32B](https://img.shields.io/badge/Model-32B-green?logo=huggingface)](https://huggingface.co/Qwen/WebWorld-32B)
 ## 📚 Introduction
@@ -44,13 +54,13 @@ Agents trained on WebWorld-synthesized trajectories achieve **+9.9% on MiniWob++
 ## 🎯 Model Series
-| Model | Base Model | Link |
-|---|---|---|
-| **WebWorld-8B** | [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) | [🤗 HuggingFace](https://huggingface.co/Qwen/WebWorld-8B) |
-| **WebWorld-14B** | [Qwen3-14B](https://huggingface.co/Qwen/Qwen3-14B) | [🤗 HuggingFace](https://huggingface.co/Qwen/WebWorld-14B) |
-| **WebWorld-32B** | [Qwen3-32B](https://huggingface.co/Qwen/Qwen3-32B) | [🤗 HuggingFace](https://huggingface.co/Qwen/WebWorld-32B) |
-**Dataset**: [Qwen/WebWorldData](https://huggingface.co/datasets/Qwen/WebWorldData)
 💡 **Recommendation**: Use 8B for fast simulation and data synthesis; use 14B/32B for higher-fidelity simulation and better long-horizon robustness. For best results in a specific environment, we recommend task-specific fine-tuning on in-domain trajectories.

 # WebWorld 🌐
+[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/LICENSE-2.0)
+[![GitHub](https://img.shields.io/badge/GitHub-WebWorld-4b32c3?logo=github)](https://github.com/QwenLM/WebWorld)
+[![Dataset](https://img.shields.io/badge/HF%20Dataset-WebWorldData-yellow?logo=huggingface)](https://huggingface.co/datasets/Qwen/WebWorldData)
+[![MS Dataset](https://img.shields.io/badge/ModelScope-Dataset-7B42BC)](https://modelscope.cn/datasets/Qwen/WebWorldData)
+[![8B](https://img.shields.io/badge/Model-8B-green?logo=huggingface)](https://huggingface.co/Qwen/WebWorld-8B)
+[![MS 8B](https://img.shields.io/badge/ModelScope-8B-7B42BC)](https://modelscope.cn/models/Qwen/WebWorld-8B)
+[![14B](https://img.shields.io/badge/Model-14B-green?logo=huggingface)](https://huggingface.co/Qwen/WebWorld-14B)
+[![MS 14B](https://img.shields.io/badge/ModelScope-14B-7B42BC)](https://modelscope.cn/models/Qwen/WebWorld-14B)
+[![32B](https://img.shields.io/badge/Model-32B-green?logo=huggingface)](https://huggingface.co/Qwen/WebWorld-32B)
+[![MS 32B](https://img.shields.io/badge/ModelScope-32B-7B42BC)](https://modelscope.cn/models/Qwen/WebWorld-32B)
 ## 📚 Introduction
 ## 🎯 Model Series
+| Model | Base Model | HuggingFace Link | ModelScope Link |
+|---|---|---|---|
+| **WebWorld-8B** | [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) | [🤗 HuggingFace](https://huggingface.co/Qwen/WebWorld-8B) | [🤖 ModelScope](https://modelscope.cn/models/Qwen/WebWorld-8B) |
+| **WebWorld-14B** | [Qwen3-14B](https://huggingface.co/Qwen/Qwen3-14B) | [🤗 HuggingFace](https://huggingface.co/Qwen/WebWorld-14B) | [🤖 ModelScope](https://modelscope.cn/models/Qwen/WebWorld-14B) |
+| **WebWorld-32B** | [Qwen3-32B](https://huggingface.co/Qwen/Qwen3-32B) | [🤗 HuggingFace](https://huggingface.co/Qwen/WebWorld-32B) | [🤖 ModelScope](https://modelscope.cn/models/Qwen/WebWorld-32B) |
+**WebWorldData**: [Huggingface: Qwen/WebWorldData](https://huggingface.co/datasets/Qwen/WebWorldData), [ModelScope: Qwen/WebWorldData](https://modelscope.cn/datasets/Qwen/WebWorldData)
 💡 **Recommendation**: Use 8B for fast simulation and data synthesis; use 14B/32B for higher-fidelity simulation and better long-horizon robustness. For best results in a specific environment, we recommend task-specific fine-tuning on in-domain trajectories.