Ukrainian platform Lapathoniia has launched Rukopys OCR, a service that recognizes handwritten Ukrainian text and converts it into digital format. The tool works through a web editor and an API, with its infrastructure hosted in Ukraine on servers run by De Novo, according to a report from De Novo.
Rukopys OCR can process full pages of documents in JPEG, PNG, WebP, TIFF and PDF formats. Once the text is recognized, users can edit it and save the result as a TXT file, JSON file or searchable PDF.
How it compares to Gemini 3 Flash
According to the developers, Rukopys OCR outperformed Gemini 3 Flash during testing. Overall accuracy reached 0.9137, compared with 0.7838 for the Google model. The Ukrainian system made 64% fewer errors when recognizing entire pages and 59% fewer errors on individual text blocks.
Gemini is part of Google's family of AI models, used in tools ranging from chatbots to document analysis. Optical character recognition, or OCR, is the technology that lets software read text from scanned images or photographs and turn it into editable digital text, a task that becomes far harder when the source is handwritten rather than printed.
Pricing and licensing
Rukopys OCR is currently available for free testing. For large-scale processing, the developers set a price of about $0.03 per page in standard mode and $0.018 per page in batch mode.
The underlying model, called Rukopys-OCR-4B, was built on top of Qwen3.5-4B and released under the Apache 2.0 license. Qwen is an open-weight AI model series developed by Alibaba, and Apache 2.0 is a widely used open-source license that allows others to freely use, modify and distribute the software.
Part of a wider push in Ukrainian AI
The launch follows other recent moves to build Ukrainian-language AI tools. Ukraine previously launched a ChatGPT-style assistant built on Ukrainian AI models, and a separate Ukrainian large language model is being trained on the basis of Google's Gemma model.
