Qwen3-32B-AWorld/README.md

---
license: apache-2.0
base_model:
- Qwen/Qwen3-32B
library_name: transformers
---

<style>
  .no-border-table table, .no-border-table th, .no-border-table td {
    border: none !important;
  }
</style>

<div class="no-border-table">

| | |
|-|-|
| [![GitHub](https://img.shields.io/badge/GitHub-Repository-black?logo=github)](https://github.com/inclusionAI/AWorld/tree/main/train) | [![arXiv](http://img.shields.io/badge/cs.AI-arXiv%3A2508.20404-B31B1B.svg?logo=arxiv&logoColor=red)](https://arxiv.org/abs/2508.20404) |

</div>

# Qwen3-32B-AWorld

## Model Description

**Qwen3-32B-AWorld** is a large language model fine-tuned from `Qwen3-32B`, specializing in agent capabilities and proficient tool usage. The model excels at complex agent-based tasks through precise integration with external tools, achieving a pass@1 score on the GAIA benchmark that surpasses GPT-4o and is comparable to DeepSeek-V3.

<img src="qwen3-32b-aworld-performance.jpg" style="width:100%;">

## Quick Start

This guide provides instructions for quickly deploying and running inference with `Qwen3-32B-AWorld` using vLLM.

### Deployment with vLLM

To deploy the model, use the following `vllm serve` command:

```bash
vllm serve inclusionAI/Qwen3-32B-AWorld \
--rope-scaling '{"rope_type":"yarn","factor":4.0,"original_max_position_embeddings":32768}' \
--max-model-len 131072 \
--gpu-memory-utilization 0.85 \
--dtype bfloat16 \
--tensor-parallel-size 8 \
--enable-auto-tool-choice \
--tool-call-parser hermes
```

**Key Configuration:**

*   **Deployment Recommendation:** We recommend deploying the model on **8 GPUs** to enhance concurrency. The `tensor-parallel-size` argument should be set to the number of GPUs you are using (e.g., `8` in the command above).
*   **Tool Usage Flags:** To enable the model's tool-calling capabilities, it is crucial to include the `--enable-auto-tool-choice` and `--tool-call-parser hermes` flags. These ensure that the model can correctly process tool calls and parse the results.

### Making Inference Calls

When making an inference request, you must include the `tools` you want the model to use. The format should follow the official OpenAI API specification.

Here is a complete Python example for making an API call to the deployed model using the requests library. This example demonstrates how to query the model with a specific tool.

```python
import requests
import json

# Define the tools available for the model to use
tools = [
    {
      "type": "function",
      "function": {
        "name": "mcp__google-search__search",
        "description": "Perform a web search query",
        "parameters": {
          "type": "object",
          "properties": {
            "query": {
              "description": "Search query",
              "type": "string"
            },
            "num": {
              "description": "Number of results (1-10)",
              "type": "number"
            }
          },
          "required": [
            "query"
          ]
        }
      }
    }
]

# Define the user's prompt
messages = [
    {
        "role": "user", 
        "content": "Search for hangzhou's weather today."
    }
]

# Set generation parameters
temperature = 0.6
top_p = 0.95
top_k = 20
min_p = 0

# Prepare the request payload
data = {
    "messages": messages,
    "tools": tools,
    "temperature": temperature,
    "top_p": top_p,
    "top_k": top_k,
    "min_p": min_p,
}

# The endpoint for the vLLM OpenAI-compatible server
# Replace {your_ip} and {your_port} with the actual IP address and port of your server.
url = "http://{your_ip}:{your_port}/v1/chat/completions"

# Send the POST request
response = requests.post(
    url,
    headers={"Content-Type": "application/json"},
    data=json.dumps(data)
)

# Print the response from the server
print("Status Code:", response.status_code)
print("Response Body:", response.text)

```

**Note:**

*   Remember to replace `{your_ip}` and `{your_port}` in the `url` variable with the actual IP address and port where your vLLM server is running. The default port is typically `8000`.