Reducto's r-1 model parses documents in one pass at lower cost

Reducto released r-1, a document parsing model that combines OCR, layout detection, table extraction, and formatting into a single processing pass, replacing a multi-stage pipeline. The company reports a 20% reduction in errors and a cost of about 1 cent per page, down from 3-6 cents with the previous approach.

Why it matters: Document parsing is a common bottleneck for enterprise AI pipelines that need to turn PDFs and scans into clean, structured text for retrieval or fine-tuning. A single-pass model that is both more accurate and several times cheaper could shift how companies build these ingestion pipelines, away from stitched-together OCR-plus-LLM workflows toward specialized end-to-end models.