Introducing r-1 and our new generation of document parsing models

Today we're announcing r-1, the first in a new series of models that sets a higher standard for document parsing. It’s more accurate than our most powerful agentic OCR models, faster, and up to 6x cheaper.

Reducto's Agentic OCR already leads the industry on accuracy. With r-1, we're making the best even better.

One model for all of your parsing tasks 

Documents capture the world’s most important data, but the files that matter most often come in forms that make them challenging to ingest. Financial statements with nested tables, contracts where a strikethrough changes the meaning, dense layouts that carry information in their structure. Getting these right, reliably and at scale, is what separates a parser you can build on from one you can't.

Reducto’s Parse has led the industry for the long tail where other pipelines fall short. We want to go further.

We spent the last few months iterating on a completely new architecture for our parsing pipeline, training it on some of the hardest document data in the world, with the goal of handling complex layouts, tables, formatting, and citations in a single pass rather than across multiple model calls.

r-1 is the first in an upcoming series of models built on that architecture. It preserves the things customers rely on — accurate layout detection, intelligent reading order, and consistent output across document types — while pushing capabilities further on the hardest cases: strikethroughs, watermarked content, dense tables, and documents that don't follow predictable templates.
Our early preview already outperforms our legacy agentic pipelines with a 20% reduction in error rate, and we'll continue iterating toward the goal of near-perfect accuracy on even the most challenging documents.

Transparent pricing that scales with you

Accuracy is only half of the story. Our legacy agentic models ran between 3 and 6 cents per page depending on the workload. r-1 costs 1 cent per page, all in.

This is part of a broader shift in how we think about pricing at Reducto. We believe transparent pricing is honest pricing, and our team is simplifying all endpoints into clear product rate cards that make it easier for customers to understand and forecast their costs. r-1 is a good example of that in practice: one price for the end-to-end parsing, without hidden multipliers and credit costs to reach high accuracy.

Later this year we'll also be releasing r-1 mini, a smaller variant designed for workloads where speed and cost matter most, alongside automatic routing that selects the right model for each page. The goal is the best result at the right price, with no orchestration overhead for you.

Try it on your own documents

r-1 is available today in preview mode for general use, with a specific configuration flag in your Parse API call.

If you’re currently using another document parser, we’ll provide up to $5,000 in credits and work with your team to make it easy to run a side-by-side comparison and migrate your workload.

We’re excited to see what you’ll build with it.