Reducto

Reducto has launched r-1, a document parsing model that converts complex PDFs, scans, spreadsheets and other files into structured data for AI systems. It is aimed at teams whose pipelines break on the long tail of real-world documents, including dense tables that span multiple pages, handwriting, low-quality scans, watermarked content, checkboxes, strikethroughs and unusual layouts. According to Reducto, r-1 reduces errors by up to 20%, improves latency at high volumes, and brings the all-in cost down to 1¢ per page , folding layout detection, reading order, table extraction, formatting and bounding box citations into a single request. The launch matters because production document AI is still typically a stitched-together stack of hyperscaler OCR, vision language models and custom post-processing. r-1 combines layout detection, reading order, tables, formatting, grounding, and granular citations in a single model , priced flat so teams can point high-volume workloads at one endpoint instead of routing pages between providers. It is the first in a planned family, with r-1 mini, a smaller model for speed- and cost-sensitive workloads, and automatic routing that selects the right model for each page on the roadmap. Teams already using another parser can request up to $5,000 in credits to benchmark r-1 against their existing setup. Reducto was founded by Adit Abraham and Raunak Chowdhuri out of MIT , with Abraham as CEO and Chowdhuri as CTO. The company has raised more than $108 million from investors including Andreessen Horowitz and First Round Capital and has processed more than a billion pages a month , and counts Harvey, Scale AI, Vanta, Airtable, Toast, and Fortune 10 enterprises among its customers.