Reducto Releases r-1: A Single Pass Document Parsing Model That Cuts Errors 20% at 1 Cent Per Page

Reducto has launched a single-page document parsing model called r-1, aimed at replacing the existing multi-stage proxy OCR process with a rewritten architecture. This model integrates text, tables, layouts, and reading order into a single full-page processing step, eliminating delays caused by additional model calls and reducing the error rate by 20%. The cost per page is reduced to 1 cent. Currently, r-1 is available in a preview version on Reducto-hosted V3 API; it does not support local deployment. However, the platform supports multi-tenant clouds, customer VPCs, and offline installation, and has passed SOC 2 Type II certification. Internal tests show that r-1 outperforms mainstream products like Amazon Textract and Azure Document Intelligence in performance. Its price advantage comes from the removal of functional multipliers and credit costs. The model natively supports processing of digital text, scanned documents, handwritten content, merged cells, and complex formats, and returns relative page boundaries. Since migration requires using V3…