Scans of invoices, statements, IDs and contracts are the most sensitive documents a company has, and they are exactly what tends to get uploaded to cloud OCR. Open OCR models turn a page into text, tables and structured Markdown right on your own server: documents stay inside the perimeter, an entire archive can be processed without per-page billing, and the model can be fine-tuned on your own form layouts. Modern models do more than recognize — they extract specific fields against a template, so the data lands straight in your accounting system. What you give up is the finished product: interfaces, validation and approval routing are integration work you do yourself. On poor scans, handwriting and complex multi-page tables errors persist, so an accounting workflow still needs a human review step.
A multilingual document parsing model: text, tables, formulas and reading order in one pass. dots.mocr also turns charts and diagrams into vector SVG.
A compact document parsing model from the popular PaddleOCR toolkit. Per the model card it supports 109 languages, including Russian; version 1.6 leads the OmniDocBench benchmark.
A lightweight OCR model from Zhipu for document parsing. The model card lists Russian among supported languages; built for high load and low-end hardware.
A model that converts documents to Markdown with tables, stamps, signatures, checkboxes and watermarks. The OCR2 model card lists Russian among its languages.
A compact OCR toolkit from the makers of Marker and Chandra: text recognition, page layout, reading order and tables. Surya OCR 2 (650M) also runs on a CPU; Russian scored 88.8% in benchmarks.
One of the strongest open vision models: reads documents, tables, charts and video, and works with user interfaces. Since Qwen3.5, vision is built directly into the main Qwen model.
dots.ocr, PaddleOCR-VL, GLM-OCR and Qwen3-VL are listed as commercially usable, while Surya's weights carry an AI Pubs OpenRAIL-M license with revenue and funding thresholds, and some Nanonets-OCR cards do not state a license at all. Those two should go past a lawyer before deployment.
An open model can run on your own server: data stays in-house, there is no per-request fee, and the model can be fine-tuned on your documents.
