olmOCR is an open-source tool designed for high-throughput conversion of PDFs and other documents into plain text while preserving natural reading order. It supports tables, equations, handwriting, and more.
olmOCR has been trained on academic papers, technical documentation, and other reference content, and uses a unique prompting technique to increase accuracy and decrease hallucinations. For full details on the recipe, read our technical report. The current model was fine-tuned on English documents; other languages are not likely to work.
Try the demo on your own documents below. You can then deploy the full olmOCR toolkit on your own GPUs for efficient, scalable document processing—at an estimated cost of just converted.
⚠️ This demo processes pages sequentially; for full throughput, use batch mode in our toolkit. ⚠️
Dataset and CheckpointsAnalyze any PDF, JPG, or PNG
Or try a sample document
