Reducto's r-1 Parsing Model Promises Fewer Errors at Lower Cost
Reducto's new r-1 model processes entire pages in one pass, aiming to cut document parsing errors by a fifth while charging a penny per page.
Reducto has introduced r-1, a document parsing model that handles optical character recognition, layout detection, table extraction, formatting, and grounding in a single full-page pass. The release on September 1, 2026, positions r-1 as a direct alternative to the company's existing multi-stage agentic pipeline, which breaks these tasks into separate steps.
The key selling point for potential adopters is a reported 20% reduction in parsing errors compared with the older approach. Reducto also highlights a flat pricing structure of one cent per page, in contrast to the legacy pipeline's variable cost of three to six cents per page. These figures are central to the migration decision for enterprises processing large volumes of documents.
By consolidating the parsing workflow into one stage, r-1 aims to simplify integration and reduce latency. The model is designed for use cases where accuracy and cost efficiency are critical, such as data extraction from invoices, forms, and other structured documents. Reducto did not disclose specific benchmark details or a timeline for broader availability.