Reducto

Reducto is a document-ingestion API that converts complex PDFs, scans, spreadsheets and slides into structured, citation-grounded JSON for LLM and agent pipelines. It is built for AI engineering and platform teams in finance, legal and healthcare whose agents fail when a parser misreads a table, and it removes the need to stitch together separate OCR, layout and extraction tools.

Reducto is an agentic document-processing API and workspace that turns PDFs, scans, spreadsheets and slide decks into structured, citation-grounded data that LLM applications and AI agents can consume. Founded in San Francisco by Adit Abraham (CEO) and Raunak Chowdhuri (CTO) and backed by Y Combinator, it combines vision-language models with traditional OCR and exposes a family of endpoints: Parse (layout-aware conversion to JSON), Extract (typed fields returned with bounding boxes and citations), Split (segmenting files into sub-documents), Classify (routing with plain-language taxonomies) and Edit, which writes extracted data back into finished PDF and DOCX files, plus Reducto Studio, a no-code environment for building, testing and deploying pipelines. On September 1, 2026 the company released r-1, its first in-house frontier parsing model, which handles layout detection, reading order, tables, formatting, grounding and citations in a single page pass instead of chaining multiple model calls. Reducto prices r-1 at a flat 1 cent per page, against the 3 to 6 cents its legacy agentic pipelines cost, and reports up to a 20% lower error rate, a figure measured against its own earlier system rather than an independent benchmark. r-1 is in preview behind a Parse API configuration flag; a smaller r-1 mini and automatic per-page model routing are planned. Reducto raised a $75M Series B led by Andreessen Horowitz in October 2025, bringing total funding to $108M, names Harvey, Scale AI, Vanta, Airtable and Toast among its customers, and says it processes more than one billion pages a month. It competes with hyperscaler services such as Amazon Textract, Azure Document Intelligence and Google Document AI, and with startups including Extend and Unstructured.

Category

Data & Analytics

Pricing

Usage-based, Contact for pricing

Target Market

CTOs, Heads of AI, Enterprise Developers, Data Engineers

Deployment

Cloud-first, API-based, Hybrid

Founded

2023

Headquarters

San Francisco, USA

• ✓
  r-1 single-pass parsing model

  Handles layout, reading order, tables, formatting and citations in one model pass at a flat 1 cent per page, replacing multi-call pipelines.

• ✓
  Extract with citations and bounding boxes

  Returns typed JSON fields tied to bounding boxes and source citations, so downstream agents and human reviewers can verify every value.

• ✓
  Split and Classify

  Segments long files into meaningful sub-documents and routes them with plain-language taxonomies, useful for mixed packets and bundles.

• ✓
  Edit API

  Writes extracted or generated data back into finished PDF and DOCX files, closing the loop on form-filling and document generation workflows.

• ✓
  Reducto Studio

  No-code workspace to build, test and deploy document pipelines, with per-field accuracy scoring and side-by-side result comparisons.

• ✓
  Enterprise deployment options

  VPC, on-premises and air-gapped deployments on the Enterprise tier for regulated data that cannot leave the corporate network.

• •
  Financial document ingestion for AI analysts

  Parse dense financial tables from filings and statements so research agents quote the correct figures with page-level citations.

• •
  Contract and agreement processing

  Convert contracts and redlined agreements into structured text with preserved reading order and strikethroughs for downstream legal AI review.

• •
  Claims and healthcare packets

  Split and classify mixed claims packets and extract typed fields under a BAA and zero-data-retention agreement on the Growth tier and above.

• •
  RAG over scanned archives

  Turn scanned, watermarked and handwritten document archives into grounded content for retrieval-augmented generation without building a custom OCR stack.

• •
  Automated form filling

  Use the Edit API to populate PDF and DOCX forms from extracted data, removing manual re-keying in back-office workflows.

Standard is pay-as-you-go with $150 in free usage: r-1 Parse costs $10 per 1,000 pages, Extract and Split $20, Deep Extract and Deep Split $40, Classify $7.50, and Edit $60 (or $15 for pre-filled pages) per 1,000 pages. Growth and Enterprise use custom volume rates; BAAs and zero-data-retention agreements start at Growth, while VPC/on-prem deployment, SSO/SAML, RBAC and custom SLAs are Enterprise-only. At the r-1 launch Reducto offered up to $5,000 in credits to teams benchmarking it on their hardest documents.

This page was written from 9 sources, 6 on domains other than reducto.ai.

1. 1.reducto.ai — parse r 1 modelvendor
2. 2.reducto.ai — pricingvendor
3. 3.reducto.ai — reducto.aivendor
4. 4.prnewswire.com — reducto unveils a frontier parsing model that makes the worl
5. 5.prnewswire.com — reducto raises 75m series b to define the future of ai docum
6. 6.marktechpost.com — reducto releases r 1 a single pass document parsing model th
7. 7.eneralabs.com — reducto r1 enterprise document ai agents 2026
8. 8.news.ycombinator.com — item
9. 9.news.ycombinator.com — item