Reducto Launches r-1 Model to Streamline Complex Document Parsing

Most AI systems stumble when documents move beyond simple text, struggling with messy handwriting, overlapping tables, or strikethroughs that fundamentally alter legal or financial meaning. While cloud giants like Amazon and Microsoft have standardized basic processing, organizations currently rely on complex, stitched-together architectures to handle high-fidelity data. Reducto claims its new model outperforms these established hyperscaler products by reducing parsing errors by up to 20%.

By unifying grounding, citations, and formatting detection, r-1 eliminates the need for expensive, layered post-processing. The model is built specifically for the "long tail" of difficult files—low-quality scans, watermarked content, and non-standard layouts—that typically cause conventional parsers to fail. To encourage adoption, the company is offering up to $5,000 in migration credits to organizations willing to benchmark their most difficult document sets against the new software. Looking ahead, Reducto plans to expand the family with a lightweight 'r-1 mini' variant designed for speed-sensitive workloads, further diversifying its agentic document platform.