---
title: WebDancer-32B
canonical_url: "https://www.modelscope.cn/models/iic/WebDancer-32B"
md_url: "https://www.modelscope.cn/models/iic/WebDancer-32B.md"
repository: iic/WebDancer-32B
last_updated: 2025-09-28
license: mit
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - qwen2
architectures:
  - Qwen2ForCausalLM
base_model:
  - Qwen/QwQ-32B
base_model_relation: finetune
parameters: 32.8B
tensor_type:
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - Pytorch
inference_backends:
  - "deploy_task text/emb"
  - "lmdeploy 0.9.1"
  - "lmdeploy_turbomind 0.9.1"
  - "sglang 0.5.2"
  - "vllm 0.9.2"
downloads: 433
stars: 3
---

# WebDancer-32B

> WebDancer-32B - iic 在 ModelScope 开源的模型。This model was presented in the paper WebDancer: Towards Autonomous Information Seeking Agency.

iic/WebDancer-32B 是 ModelScope 魔搭社区上的 32.8B 参数text-generation模型，采用 mit 许可，基于 Qwen/QwQ-32B 构建，可用 deploy_task text/emb、lmdeploy 0.9.1、lmdeploy_turbomind 0.9.1 部署。

- **Repository**: iic/WebDancer-32B
- **License**: mit
- **Tasks**: text-generation
- **Parameters**: 32.8B
- **Base model**: Qwen/QwQ-32B
- **Inference backends**: deploy_task text/emb, lmdeploy 0.9.1, lmdeploy_turbomind 0.9.1, sglang 0.5.2, vllm 0.9.2
- **Downloads**: 433
- **Stars**: 3
- **Last updated**: 2025-09-28

Source: https://www.modelscope.cn/models/iic/WebDancer-32B

---

This model was presented in the paper [WebDancer: Towards Autonomous Information Seeking Agency](https://huggingface.co/papers/2505.22648).

You can download the model then run the inference scipts in https://github.com/Alibaba-NLP/WebAgent.

- Native agentic search reasoning model using ReAct framework towards autonomous information seeking agency and Deep Research-like model.
- We introduce a four-stage training paradigm comprising browsing data construction, trajectory sampling, supervised fine-tuning for effective cold start, and reinforcement learning for improved generalization, enabling the agent to autonomously acquire autonomous search and reasoning skills.
- Our data-centric approach integrates trajectory-level supervision fine-tuning and reinforcement learning (DAPO) to develop a scalable pipeline for training agentic systems via SFT or RL.
- WebDancer achieves a Pass@3 score of 61.1% on GAIA and 54.6% on WebWalkerQA.