Home › RAG and Search › Tongyi DeepResearch

Tongyi DeepResearch - AI RAG and Search Tool

Overview

Tongyi DeepResearch is an open-source research agent and specialized 30B MoE (Mixture-of-Experts) language model developed by Alibaba Tongyi Lab, designed for long-horizon web research workflows. The project emphasizes agent-style reasoning and web-scale information gathering, combining model weights, inference-mode workflows, and tooling to support iterative research tasks that require browsing, multi-step planning, and document-level reading. According to the GitHub repository (https://github.com/Alibaba-NLP/DeepResearch), the release includes model checkpoints on Hugging Face/ModelScope, live demos, and reference code for running the agent locally. The codebase supports multiple inference paradigms—explicitly ReAct and IterResearch—so agents can interleave actions (search, page reading) and reasoning over extended sessions. Tongyi DeepResearch also bundles evaluation scripts and integrations for search and page-reading, enabling users to build Retrieval-Augmented Generation (RAG) pipelines and deploy the agent for web research, automated literature review, or multi-step question answering. The repository is positioned as both a model release (30B MoE weights) and a practical toolkit for researchers and engineers to reproduce and extend web-based agent behaviors.

Key Features

  • Specialized 30B Mixture-of-Experts (MoE) language model for long-horizon research.
  • Supports ReAct inference mode for interleaved reasoning and external actions.
  • IterResearch mode for iterative, multi-step research workflows and follow-up queries.
  • Provides Hugging Face and ModelScope weights plus live online demos for testing.
  • Reference code for agent browsing, search/page-reading integration, evaluation, deployment.

Community

Tongyi DeepResearch is published as an open-source project on GitHub, with model weights surfaced via Hugging Face and ModelScope. The repository provides demos, issue tracking, and code contributions pathways (issues/PRs) for community feedback and extension. For up-to-date activity, contributions, and release notes, consult the project repository at https://github.com/Alibaba-NLP/DeepResearch.

Last Refreshed: 2026-01-09

Key Information

  • Category: RAG and Search
  • Type: AI RAG and Search Tool
Visit Official Website

Related Tools

Crawl4AI

An open-source, LLM-friendly tool designed to crawl and extract data, facilitating content aggregation for AI applications.
Perplexica

Open-source AI-powered search engine that performs deep web research to generate and cite answers.
GPT Researcher

An LLM-based autonomous agent that conducts deep local and web research on any topic and generates long reports with citations, with support for connecting to specialized data sources.
Maxun

An open-source no-code web data extraction platform that lets users train a robot in minutes to automatically scrape websites and convert them into APIs and spreadsheets.
Auto-Deep-Research

An open-source, fully-automated personal AI assistant that serves as a cost-effective alternative to OpenAI's Deep Research. Built on the AutoAgent framework, it supports integration with various LLMs, function-calling interactions, file uploads, and a one-click launch for effortless research automation.
Open Deep Research Agent

An open–source deep research AI agent that utilizes reasoning models to conduct in–depth factual research. It separates planning and research execution for detailed report generation.