---
language:
- en
library_name: transformers
license: apache-2.0
pipeline_tag: text-generation
---
# Tongyi DeepResearch: WebWeaver
This repository contains **Tongyi DeepResearch**, an agentic large language model, which is associated with the paper [WebWeaver: Structuring Web-Scale Evidence with Dynamic Outlines for Open-Ended Deep Research](https://huggingface.co/papers/2509.13312).
[](https://huggingface.co/papers/2509.13312) |
[](https://github.com/Alibaba-NLP/DeepResearch) |
[](https://tongyi-agent.github.io/blog/introducing-tongyi-deep-research/)
## Introduction
We present **Tongyi DeepResearch**, an agentic large language model featuring 30 billion total parameters, with only 3 billion activated per token. Developed by Tongyi Lab, the model is specifically designed for **long-horizon, deep information-seeking** tasks. Tongyi-DeepResearch demonstrates state-of-the-art performance across a range of agentic search benchmarks, including Humanity's Last Exam, BrowserComp, BrowserComp-ZH, WebWalkerQA, GAIA, xbench-DeepSearch and FRAMES.
More details can be found in our 📰 [Tech Blog](https://tongyi-agent.github.io/blog/introducing-tongyi-deep-research).
## Key Features
- ⚙️ **Fully automated synthetic data generation pipeline**: We design a highly scalable data synthesis pipeline, which is fully automatic and empowers agentic pre-training, supervised fine-tuning, and reinforcement learning.
- 🔄 **Large-scale continual pre-training on agentic data**: Leveraging diverse, high-quality agentic interaction data to extend model capabilities, maintain freshness, and strengthen reasoning performance.
- 🔁 **End-to-end reinforcement learning**: We employ a strictly on-policy RL approach based on a customized Group Relative Policy Optimization framework, with token-level policy gradients, leave-one-out advantage estimation, and selective filtering of negative samples to stabilize training in a non‑stationary environment.
- 🤖 **Agent Inference Paradigm Compatibility**: At inference, Tongyi-DeepResearch is compatible with two inference paradigms: ReAct, for rigorously evaluating the model's core intrinsic abilities, and an IterResearch-based 'Heavy' mode, which uses a test-time scaling strategy to unlock the model's maximum performance ceiling.
# Model Download
You can directly download the model by following the links below.
| Model | Download Links | Model Size | Context Length |
| :-----------------: | :-----------------------------------------: | :----------: | :--------------: |
| Tongyi-DeepResearch-30B-A3B | [🤗 HuggingFace](https://huggingface.co/Alibaba-NLP/Tongyi-DeepResearch-30B-A3B)
[🤖 ModelScope](https://modelscope.cn/models/iic/Tongyi-DeepResearch-30B-A3B) | 30B-A3B | 128K |
# News
[2025/09/17]🔥 We have released **Tongyi-DeepResearch-30B-A3B**.
# Deep Research Benchmark Results
## Quick Start
This guide provides instructions for setting up the environment and running inference scripts located in the [inference](https://github.com/Alibaba-NLP/DeepResearch/tree/main/inference/) folder.
### 1. Environment Setup
- Recommended Python version: **3.10.0** (using other versions may cause dependency issues).
- It is strongly advised to create an isolated environment using `conda` or `virtualenv`.
```bash
# Example with Conda
conda create -n react_infer_env python=3.10.0
conda activate react_infer_env
```
### 2. Installation
Install the required dependencies:
```bash
pip install -r requirements.txt
```
### 3. Prepare Evaluation Data
- Create a folder named `eval_data/` in the project root.
- Place your QA file in **JSONL** format inside this directory, e.g. `eval_data/example.jsonl`.
- Each line must be a JSON object that includes **both** of the following keys:
```json
{"question": "...","answer": "..."}
```
- A sample file is provided in the `eval_data` folder for reference.
- If you plan to use the *file parser* tool, **prepend the file name to the `question` field** and place the referenced file inside the `eval_data/file_corpus/` directory.
### 4. Configure the Inference Script
- Open `run_react_infer.sh` and modify the following variables as instructed in the comments:
* `MODEL_PATH` - path to the local or remote model weights.
* `DATASET` - path to the evaluation set, e.g. `example`.
* `OUTPUT_PATH` - path for saving the prediction results, e.g. `./outputs`.
- Depending on the tools you enable (retrieval, calculator, web search, etc.), provide the required `API_KEY`, `BASE_URL`, or other credentials. Each key is explained inline in the bash script.
### 5. Run the Inference Script
```bash
bash run_react_infer.sh
```
---
With these steps, you can fully prepare the environment, configure the dataset, and run the model. For more details, consult the inline comments in each script or open an issue.
## Benchmark Evaluation
We provide benchmark evaluation scripts for various datasets. Please refer to the [evaluation scripts](https://github.com/Alibaba-NLP/DeepResearch/tree/main/evaluation/) directory for more details.
## Deep Research Agent Family
Tongyi DeepResearch also has an extensive deep research agent family. You can find more information in the following paper:
[1] [WebWalker: Benchmarking LLMs in Web Traversal](https://arxiv.org/pdf/2501.07572)
[2] [WebDancer: Towards Autonomous Information Seeking Agency](https://arxiv.org/pdf/2505.22648)
[3] [WebSailor: Navigating Super-human Reasoning for Web Agent](https://arxiv.org/pdf/2507.02592)
[4] [WebShaper: Agentically Data Synthesizing via Information-Seeking Formalization](https://arxiv.org/pdf/2507.15061)
[5] [WebWatcher: Breaking New Frontier of Vision-Language Deep Research Agent](https://arxiv.org/pdf/2508.05748)
[6] [WebResearcher: Unleashing unbounded reasoning capability in Long-Horizon Agents](https://arxiv.org/pdf/2509.13309)
[7] [ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization](https://arxiv.org/pdf/2509.13313)
[8] [WebWeaver: Structuring Web-Scale Evidence with Dynamic Outlines for Open-Ended Deep Research](https://arxiv.org/pdf/2509.13312)
[9] [WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning](https://arxiv.org/pdf/2509.13305)
[10] [Scaling Agents via Continual Pre-training](https://arxiv.org/pdf/2509.13310)
[11] [Towards General Agentic Intelligence via Environment Scaling](https://arxiv.org/pdf/2509.13311)
## 🌟 Misc
[](https://www.star-history.com/#Alibaba-NLP/DeepResearch&Date)
## 🚩 Talent Recruitment
🔥🔥🔥 We are hiring! Research intern positions are open (based in Hangzhou、Beijing、Shanghai)
📚 **Research Area**：Web Agent, Search Agent, Agent RL, MultiAgent RL, Agentic RAG
☎️ **Contact**：[yongjiang.jy@alibaba-inc.com]()
## Contact Information
For communications, please contact Yong Jiang (yongjiang.jy@alibaba-inc.com).
## Citation
```bibtex
@misc{webweaver2025,
title={WebWeaver: Structuring Web-Scale Evidence with Dynamic Outlines for Open-Ended Deep Research},
author={Tongyi DeepResearch Team},
year={2025},
eprint={2509.13312},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://huggingface.co/papers/2509.13312}
}
@misc{tongyidr,
author={Tongyi DeepResearch Team},
title={Tongyi-DeepResearch},
year={2025},
howpublished={\url{https://github.com/Alibaba-NLP/DeepResearch}}
}
```