# Alibaba-NLP/DeepResearch

Tongyi Deep Research, the Leading Open-source Deep Research Agent

Repository: https://github.com/Alibaba-NLP/DeepResearch
Canonical: https://ross.abutalabs.com/products/deepresearch
Homepage: https://tongyi-agent.github.io/blog/introducing-tongyi-deep-research/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: agent, llm, artificial-intelligence, information-seeking, alibaba, web-agent, deepresearch, tongyi, deep-research
Last push: 2026-02-27T13:12:11+00:00

## Health v2 (maintenance only)
Score: 52/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 69, release rhythm 35, longevity 42
- inputs: {"age_days": 601, "days_push": 187, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 19878, forks 1516 (observed 2026-08-28T04:11:29.347721+00:00)

## What it is
Tongyi DeepResearch is an open-source agentic large language model (30.5B total parameters, 3.3B activated per token) plus training and inference code, designed for long-horizon, deep information-seeking tasks on the web. It includes the model weights, a ReAct inference framework, a test-time-scaling Heavy Mode, and a full synthetic-data training pipeline spanning continual pre-training, SFT, and on-policy agent reinforcement learning.

## Use cases
- build an autonomous deep research agent that browses the web and compiles reports
- run a self-hosted alternative to OpenAI DeepResearch for complex information-seeking
- answer hard multi-hop research questions like BrowseComp or Humanity's Last Exam
- fine-tune or reproduce an agentic LLM with reinforcement learning and synthetic data
- deploy a research assistant that searches, reads, and synthesizes web sources
- evaluate web agents on deep search benchmarks

## When to choose
- you need an open-weights model purpose-built for long-horizon web research and information seeking
- you want to self-host a deep research agent instead of relying on proprietary APIs
- you are researching agentic LLM training pipelines (CPT, SFT, on-policy RL) and want a full reproducible methodology
- you need state-of-the-art open-source performance on deep search benchmarks

## When to avoid
- you need a lightweight chatbot for simple Q&A without web research
- you cannot host a 30B-parameter model or afford inference costs for agentic rollouts
- you need a production-ready managed service with guaranteed latency rather than a research codebase
- your task is domain-specific reasoning that does not require web information seeking

## Facets
- artifact type: library
- maturity: active
- function: agent-framework, llm-inference, rag, web-scraping, machine-learning
- domain: artificial-intelligence, large-language-models
- platform: python, self-hosted
- tags: deep-research, web-agent, information-seeking, agentic-llm, reinforcement-learning, synthetic-data, tongyi, open-weights, ai-agents, natural-language-processing, search, linux, macos, docker

## Member repositories
- Alibaba-NLP/DeepResearch (main) score 52

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:29.347721+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:59:24.493319+00:00, confidence not recorded.
  - readme: https://github.com/Alibaba-NLP/DeepResearch (fetched 2026-08-28T04:11:29.347721+00:00, sha f3fbfdb2e53a)
  - homepage: https://tongyi-agent.github.io/blog/introducing-tongyi-deep-research/ (fetched 2026-08-29T07:57:59.924159+00:00, sha 655f2611a772)
- Data as of 2026-08-30T08:39:29.467469+00:00.