Verified project record

Alibaba-NLP/DeepResearch

An open-source agentic language model designed for long-horizon, deep information-seeking tasks across the web. It features a 30.5B parameter mixture-of-experts architecture, activating 3.3B parameters per token, and supports dual inference paradigms.

Project overview

Project type: AI Agent · RAG · Model Runtime
Use cases: Search & Research
Deployment: Refer to project documentation
License: Apache-2.0

Best for

AI engineers and researchers looking to automate long-horizon, deep information-seeking tasks using an open-source agentic model with state-of-the-art benchmark performance.

Key capabilities

A 30.5B parameter MoE model designed for long-horizon, deep information-seeking tasks.
Supports ReAct for evaluating intrinsic abilities and IterResearch 'Heavy' mode using test-time scaling for maximum performance.
Capable of parsing and referencing uploaded document files like PDFs and Excel sheets during research queries.
Provides scripts to evaluate the model on various deep research and agentic search benchmarks.

Limitations and risks

Online demos may experience latency or fail intermittently.
External tool QPS (queries per second) limits may affect response times during research tasks.

Getting started

Create a conda environment with Python 3.10.0 and install dependencies using pip. Configuring API keys in a .env file is required.
Download the model weights and execute the inference script using bash via the command-line interface.

Evidence and sources

README: 👏 Welcome to try Tongyi DeepResearch via our **[<img src="./assets/tongyi.png" width="14px" style="display:inline;"> Modelscope online demo](https://www.modelscope.cn/studios/jial…
README: | Tongyi-DeepResearch-30B-A3B | [🤗 HuggingFace](https://huggingface.co/Alibaba-NLP/Tongyi-DeepResearch-30B-A3B)<br> [🤖 ModelScope](https://modelscope.cn/models/iic/Tongyi-DeepRese…
README: an agentic large language model featuring 30.5 billion total parameters, with only 3.3 billion activated per token. Developed by Tongyi Lab, the model is specifically designed for…
README: [2025/09/17]🔥 We have released **Tongyi-DeepResearch-30B-A3B**.
README: Agent Inference Paradigm Compatibility: At inference, Tongyi DeepResearch is compatible with two inference paradigms: ReAct, for rigorously evaluating the model's core intrinsic a…

Reviewed at: 2026-07-20T13:01:33Z