# RUC-NLPIR/WebThinker

[NeurIPS 2025] 🌐 WebThinker: Empowering Large Reasoning Models with Deep Research Capability

Repository: https://github.com/RUC-NLPIR/WebThinker
Canonical: https://ross.abutalabs.com/products/webthinker
Homepage: https://foremost-beechnut-8ed.notion.site/WebThinker-Empowering-Large-Reasoning-Models-with-Deep-Research-Capability-d13158a27d924a4b9df7f9ab94066b64
Language: Python
License: MIT
License Family: permissive
Topics: deepresearch, deepsearch, deepseek-r1, gaia, gpqa, hle, o1, o3, qwq, reasoning, reportgen, webwalker, research
Last push: 2025-12-08T03:40:49+00:00

## Health v2 (maintenance only)
Score: 45/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 56, release rhythm 35, longevity 37
- inputs: {"age_days": 523, "days_push": 268, "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 1467, forks 141 (observed 2026-08-28T04:04:48.760487+00:00)

## What it is
WebThinker is a research framework that empowers large reasoning models (like DeepSeek-R1 and QwQ) with deep research capability, letting them autonomously search, click, and extract information from the web while reasoning. It also provides fine-tuned model checkpoints and report generation for complex research questions.

## Use cases
- run deep research with open-source reasoning models
- generate research reports automatically from web sources
- let an LLM search and browse the web while reasoning
- evaluate reasoning models on GAIA, GPQA, HLE, or WebWalker benchmarks
- fine-tune DeepSeek-R1 or QwQ models for deep search tasks
- build an open-source alternative to proprietary deep research agents

## When to choose
- you want an open-source deep research pipeline built on large reasoning models
- you need reproducible research code from a peer-reviewed NeurIPS paper
- you want to benchmark or fine-tune R1/QwQ models on deep search tasks

## When to avoid
- you need a production-ready hosted research assistant with no setup
- you don't have GPU resources to run large reasoning models locally
- you only need simple web scraping without LLM reasoning

## Facets
- artifact type: library
- maturity: active
- function: rag, agent-framework, llm-inference, search-engine, nlp
- domain: large-language-models
- platform: python, cross-platform
- tags: deep-research, deep-search, reasoning-models, report-generation, web-search, deepseek-r1, qwq, neurips-2025, retrieval-augmented-generation, ai-agents, natural-language-processing, research, gpu

## Member repositories
- RUC-NLPIR/WebThinker (main) score 45

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:48.760487+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 2026-08-30T04:34:58.307607+00:00, confidence not recorded.
  - readme: https://github.com/RUC-NLPIR/WebThinker (fetched 2026-08-28T04:04:48.760487+00:00, sha 26bda56e59c8)
  - homepage: https://foremost-beechnut-8ed.notion.site/WebThinker-Empowering-Large-Reasoning-Models-with-Deep-Research-Capability-d13158a27d924a4b9df7f9ab94066b64 (fetched 2026-08-29T11:43:03.574244+00:00, sha 73a6ba54b760)
- Data as of 2026-08-30T08:39:29.467469+00:00.