Reducto is launching r-1 mini so AI teams can parse documents at $1 per 1,000 pages, a rate aimed at workloads that were too expensive to run through higher-accuracy parsers. The San Francisco company, cofounded by Adit Abraham and Raunak Chowdhuri, already sells r-1 at $10 per 1,000 pages. Enterprise customers were still skipping lower-priority documents at that price, so r-1 mini is meant for archives and high-volume pipelines where cost had been the limit. r-1 and r-1 mini both read a page in one pass, handling text, tables, figures, layout, formatting, and bounding boxes that link each block back to its position. That work used to be split across OCR, layout detection, and extra agentic steps. r-1 is now generally available, with word-level citations and latency Reducto says is 30 to 50 percent better than r-1 mini. The smaller model is for large, cost-sensitive jobs such as web crawling, research agents, and enterprise knowledge search. In Reducto's ParseArena evaluation of 2,000 public pages across seven industries, r-1 mini led every parser the company tested under $10 per 1,000 pages by 7 points or more. The five higher-scoring models all cost at least 10 times as much, and Reducto says r-1 mini beat options priced up to 25 times higher, including LlamaParse Agentic and Extend Performance. It also reported 60 percent fewer serious errors per page than Amazon Textract Layout + Tables, and about half as many as Azure Document Intelligence. Pages that need the most accuracy, such as charts, degraded scans, and mixed-language documents, can still be sent to r-1. Both models are available now through the Parse API.