# NuMind

> NuMind builds multimodal LLMs specialized in extracting information from documents: structured data extraction (JSON) and full-document content extraction (Markdown), without a separate OCR pipeline.

NuMind is a research company founded in June 2022. Its NuExtract model family is available via API and private deployment.

## Pages

- [Home](https://numind.ai/): NuMind builds multimodal LLMs specialized in extracting information from documents.
- [About NuMind](https://numind.ai/about): Research team, academic publications, investors and partners behind NuMind.
- [Blog](https://numind.ai/blog): Research and product notes on information extraction and document understanding.
- [Papers](https://numind.ai/papers): Peer-reviewed publications from the NuMind research team.

## Blog posts

- [NuExtract3: The Reasoning Open-Source OCR & Structured Extraction LLM](https://numind.ai/blog/nuextract-3-release): NuExtract3 is a 4B open-source VLM that unifies structured extraction (documents to JSON) and content extraction (OCR) in one reasoning model.
- [NuExtract Platform: The New Information Extraction](https://numind.ai/blog/nuextract-platform-the-new-information-extraction): The NuExtract platform extracts high-quality structured information (JSON) from documents via API, powered by NuExtract 2.0 PRO.
- [NuExtract 2.0: Outclassing Frontier LLMs in Information Extraction](https://numind.ai/blog/outclassing-frontier-llms-nuextract-2-0): NuExtract 2.0 brings vision, abstraction and in-context learning to structured extraction, outperforming GPT-4.1 and other frontier models.
- [NuExtract 1.5 - Multilingual, Infinite context, still small, and better than GPT-4o!](https://numind.ai/blog/nuextract-1-5-multilingual-infinite-context): NuExtract 1.5 is multilingual, handles arbitrarily long documents, and outperforms GPT-4o in English while being 500 times smaller.
- [NuExtract: A Foundation Model for Structured Extraction](https://numind.ai/blog/nuextract-a-foundation-model-for-structured-extraction): NuExtract is a lightweight text-to-JSON LLM that extracts arbitrarily complex information from text, open-sourced under MIT license.
- [A Foundation Model for Entity Recognition](https://numind.ai/blog/a-foundation-model-for-entity-recognition): We use modern LLMs to create a BERT-size foundation model for entity recognition, needing 5x less annotated data. MIT licensed.
- [Creating Task-Specific Foundation Models with GPT-4](https://numind.ai/blog/creating-task-specific-foundation-models-with-gpt-4): How to use LLMs to efficiently create task-specific foundation models, applied to a state-of-the-art sentiment analysis model.
- [What are Large Language Models?](https://numind.ai/blog/what-are-large-language-models): A clear introduction to large language models: what they are, how they work, and how they are created.
- [Seed Round Completed](https://numind.ai/blog/seed-round-completed): NuMind raised a $3M seed round to continue building its NLP tool focused on text understanding.

## Optional

- [SaaS Terms and Conditions](https://numind.ai/terms/saas): Terms governing access to and use of the NuExtract API services.