Yandex Vision OCR release notes

Release as of 18/05/2026Release as of 18/05/2026

Updated Vision OCR models.

Release as of 30/06/2025Release as of 30/06/2025

• Updated existing models and improved recognition accuracy.
• Added the markdown and math-markdown models that return the recognized text in Markdown syntax.
• The LayoutType property is now available for all recognized text blocks.
• The API response now includes the pictures field containing coordinates of the images recognized in the document.
• The English-Russian language model is no longer supported.

Release as of 06/02/2024Release as of 06/02/2024

• Improved table recognition quality. All tables are now properly identified as tables.
• Added Turkish language support for table recognition.

Release as of 11/01/2024Release as of 11/01/2024

Updated the table recognition model.

Release as of 07/04/2023Release as of 07/04/2023

• For document recognition, we have added the vehicle-registration-front and vehicle-registration-back models that extract data from vehicle registration certificates.
• Improved recognition accuracy for all models.
• Significantly improved PDF file processing quality.