OpenDataLab
OpenDataLab is a data-centric AI research group and open-data ecosystem focused on democratizing access to high-quality data, improving data preparation and learning efficiency, and advancing multimodal and scientific AI. Its platform unifies open datasets, data descriptions, tooling, and high-speed acquisition, while its research spans data-centric AI, multimodal models, document understanding, and AI for Science.
At a Glance
- AI researchers and data scientists
- Large-model and multimodal-model developers
- Academic and research institutions
- Developers building RAG and agent workflows
- +2 more
AI Tools by OpenDataLab
(1)
MinerU
Open Source Document Parser
Discussions
No discussions yet
Be the first to start a discussion about OpenDataLab
Latest News
MinerU 3.4 released with PP-OCRv6, about 11% OCR accuracy improvement on OmniDocBench v1.6, roughly 100% faster OCR processing, and improved model caching.
MinerU 3.1.0 released with a new Apache-2.0-based open-source license, MinerU2.5-Pro-2604-1.2B, and native PPTX/XLSX parsing.
MinerU 3.0 released with native DOCX parsing, asynchronous API tasks, mineru-router, multi-GPU deployment, and long-document scaling improvements.
Sciverse announced as a 100PB-scale AI-ready data-foundry initiative for AI for Science.
Products & Services
An open AI-data platform for discovering, filtering, visualizing, hosting, and downloading datasets across modalities and tasks. The platform uses DSDL (Data Set Description Language) to improve interoperability and reusability.
Open-source document parsing engine/command-line reader that converts PDFs, images, Office and other supported documents into AI-ready Markdown, JSON, HTML, LaTeX, DOCX, EPUB, PDF, structured content, and content-list outputs. MinerU 4.0 adds local document-library, service, SDK/API, WebUI, progressive-reading, and tiered-runtime workflows.
Intelligent data-annotation system in the OpenDataLab tool suite.
LLM-assisted intelligent annotation system for data construction.
Market Position
OpenDataLab positions itself as an open, end-to-end data-centric AI ecosystem rather than only a dataset catalog: it combines dataset hosting and discovery, a unified description language, annotation and evaluation tools, corpus construction, and document parsing. Its closest alternatives include Hugging Face Datasets, ModelScope, Kaggle, and commercial document-AI/OCR systems; MinerU differentiates through open-source, self-hosted, multimodal document parsing and structured outputs for LLM and agent workflows.
Leadership
Founders
Conghui He
Founded the OpenDataLab team at Shanghai AI Laboratory. He earned a Ph.D. at Tsinghua University, worked on 10-million-core seismic simulation and won the ACM Gordon Bell Prize, then built real-time graph-computing systems at Tencent and led a SenseTime team building data factories for more than 100,000 commercial models.
Executive Team
Conghui He
Founder and team lead, OpenDataLab; researcher at Shanghai Artificial Intelligence Laboratory
Tsinghua Ph.D.; ACM Gordon Bell Prize winner; former Tencent graph-computing engineer/leader and SenseTime data-factory team leader.
Founding Story
OpenDataLab was established in 2022 to address fragmented data sources, inconsistent formats, and the difficulty of making human knowledge usable for AI. Conghui He says the team was founded at Shanghai AI Laboratory around the conviction that data—not only algorithms or compute—was the bottleneck for foundation models; the initial vision was an open platform with unified data descriptions, intelligent search, and efficient download and processing.
Business Model
Revenue Model
The public OpenDataLab platform describes a free public data-hosting service, while the core software and datasets are open source or publicly accessible. MinerU supports local self-hosting and explicitly configured remote/API deployment; the reviewed sources do not establish a commercial pricing model for OpenDataLab itself.
Pricing Tiers
The platform paper describes a free and convenient public data-hosting service.
Target Markets
- AI researchers and data scientists
- Large-model and multimodal-model developers
- Academic and research institutions
- Developers building RAG and agent workflows
- Enterprises with document digitization and data-processing workloads
- AI for Science and scientific-research organizations
- Pre-training, fine-tuning, and evaluation data preparation for large language and multimodal models
- AI research dataset discovery and reuse
- Document digitization and conversion of PDFs, scans, Office files, and scientific literature into structured AI-ready data
- RAG and agent document reading
- Data annotation and labeling
- AI-generated-data detection and data quality assessment
- Huawei
- Alibaba