ABBYY launches FineParser, a CPU-only OCR tool for AI pipelines

FineReader repackaged for AI pipelines

ABBYY has packaged its FineReader OCR engine as a self-hosted tool for turning documents into structured text that AI systems can use. The new product, FineParser, runs in a Docker container on a CPU and does not need a GPU. Its goal is to preserve a document's layout while extracting what is inside it.

FineParser takes images of documents in multiple languages and converts them into structured, formatted text that can then be processed by generative AI models. That is not its only purpose: ABBYY says the tool can also help move printed text into a modern CMS, support archiving work, or serve as one stage in a production pipeline.

Deterministic extraction, generative output

ABBYY describes FineParser's approach as "deterministic" AI. The engine extracts text and document structure instead of generating a plausible version of them, and its output can then be handed to a generative AI system whose answers are far less predictable.

Preserving structure means recognising columns in reading order, headings and tables — including tables that have no borders — rather than returning a jumble of extracted text. ABBYY says FineParser also handles handwriting and more than 200 languages. Developers submit documents through a REST API and receive the result as plain text, JSON, or DocLang, a compact format intended as input for large language models.

Free tier, licence server, open components

FineParser itself is not open source, although ABBYY maintains a GitHub repository for examples and community support. The self-hosted tool has a free tier allowing 1,000 pages per month for one year. Subscription tiers connect to a licence server for validation, and ABBYY says a fully offline deployment requires an Enterprise plan.

The company also offers NeoML, a programmable machine-learning framework that is free and open source and published on GitHub, plus an OCR SDK for businesses that want to embed the FineReader engine in their own products.

Why it matters

The pitch behind FineParser is that before a large language model can do anything useful with a document, something has to read that document properly. ABBYY is betting that its decades-old approach to OCR — running on a CPU, preserving layout and leaving text generation to other tools — still has a place in the AI pipeline. Sometimes the old-fashioned part is the useful part.