---
id: "b928"
language: "en"
library: "tools"
official_url: "https://github.com/allenai/olmocr"
profile_updated_on: "2026-10-23"
---

# olmOCR

Convert PDFs and document images into reading-order-aware Markdown with a vision-language model.

- Company / organization: The Allen Institute for Artificial Intelligence
- Company country / region: United States · Published company address
- Category: Model development
- Official website: [https://github.com/allenai/olmocr](https://github.com/allenai/olmocr)
- Profile updated: 2026-10-23


## Overview

olmOCR linearizes PDFs, PNGs and JPEGs into Markdown for document processing at scale. Its vision-language model handles reading order across columns, tables and equations. Run local GPU inference or connect a remote vLLM server. This is a processing pipeline, not a downloadable corpus; review converted content against the source documents.

## Usage guide & costs

### Before you start

Code and the reviewed olmOCR-2-7B-1025-FP8 card state Apache 2.0. Local GPU use requires the documented NVIDIA environment and memory; remote inference sends work to the configured server and is not offline.

[Official details](https://github.com/allenai/olmocr/blob/main/README.md)
[Official details](https://github.com/allenai/olmocr/blob/main/LICENSE)
[Official details](https://huggingface.co/allenai/olmOCR-2-7B-1025-FP8)


## Official sources and references

- [Official website](https://github.com/allenai/olmocr)
- [Documentation](https://github.com/allenai/olmocr/blob/main/README.md)
- [Documentation](https://huggingface.co/allenai/olmOCR-2-7B-1025-FP8)

## Reading & subscriptions

- [Interactive profile](../../../products/olmocr--b928.html)
- [Read](../../../read/en/products/olmocr--b928.html)
- [JSON](../../../read/en/products/olmocr--b928.json)

Descriptions summarize public information; consult official websites for features, pricing and regional availability.