Reducto is a US-based document parsing company founded around 2023. It is well funded ($108M, Series B led by a16z) and well engineered, and its open RD-TableBench benchmark demonstrates real technical depth. In September 2026 it replaced its earlier multi-pass OCR-and-VLM pipeline with r-1, a single end-to-end parsing model priced flat at $10 per 1,000 pages. The product is broader than a single parsing endpoint: alongside Parse, Split, Extract and Edit it ships a Classify API and Studio, a no-code builder for chaining those steps into a pipeline.
For European enterprises, the questions that remain are jurisdictional and operational: who governs the data, which certificate the procurement team requires, and what happens to a field the model was unsure about.
Workflow builder and schema management
Reducto's Studio is a no-code builder for chaining pipeline steps, and Classify handles document typing. Describing Reducto as a bare parsing endpoint that leaves orchestration to you is out of date.
What its documentation does not describe is a human review queue: a place where a reviewer sees the fields the system was unsure about, corrects them, and the correction flows back into the run and on to downstream systems. anyformat pairs the pipeline with that review layer, and with schema management aimed at non-technical teams — an ops user uploads a sample document, sees extraction results, and changes fields without a developer and without a deployment cycle.
The question worth asking in an evaluation is not whether a builder exists, but who is able to change the extraction logic and what happens to the values nobody verified.
European sovereignty and data residency
Reducto is a US company deployed on AWS with regional endpoints for the EU and AU, and it offers VPC and on-prem deployment. That is meaningful infrastructure flexibility, but the company, its governance, and its legal framework are all US-based.
anyformat is EU-native: built by a European team, subject to European governance, GDPR-compliant by architecture. Selecting an EU endpoint on a US-governed platform does not make the data European-governed, and for regulated buyers that distinction is the whole point.
ISO 27001 and compliance
Reducto holds SOC 2 Type II and offers a HIPAA BAA. These are solid certifications for the US market. ISO 27001, the standard European procurement teams typically require, is not publicly listed among them.
anyformat is ISO 27001 certified and GDPR-compliant, with certification scope covering the document processing pipeline end to end. For European RFPs, ISO 27001 is often a hard requirement that SOC 2 alone does not satisfy.
Zero data retention
Reducto supports zero data retention account-wide, so documents are not stored after processing. On this control the two products are level.
anyformat offers zero-retention processing as a native toggle with the same outcome. What does not level out is the jurisdiction the retained-or-not data was processed under, which is where the compliance conversation actually lands.
Parse and extract capabilities
Reducto's parsing is genuinely strong. Its r-1 model reads a document end to end in a single pass rather than stitching together separate OCR, layout and correction steps, and it reports up to 20% higher accuracy than AWS, Google, and Azure on its own benchmarks, with the open RD-TableBench adding credibility. r-1 is currently a preview, available over the API only, and doesn't yet cover every configuration the legacy pipeline does.
anyformat combines LLMs with deterministic rules and calibrated confidence scoring across the full pipeline rather than the parsing step alone, adapting to 100+ formats and any layout without templates. The metric we optimize is not benchmark parsing quality but "how often is the system wrong and does not know it".
On-premise deployment
Reducto offers cloud, VPC, and fully on-prem or air-gapped deployment. anyformat offers private cloud and on-premise deployment, including air-gapped environments. Both can meet a data perimeter requirement, and neither side should win an evaluation on this point.
Accuracy, confidence and evaluation
Reducto reports 99.24% accuracy on clinical workloads (Anterior case study), publishes RD-TableBench openly, and returns a granular confidence value per extracted field. Its benchmark transparency is better than most of the market.
The difference is calibration and what sits behind the number. anyformat calibrates confidence per field, measured at 99.1% calibration accuracy with an Adaptive ECE of 0.009 (anyformat benchmark, 2026), so a threshold set for straight-through processing behaves the way the number implies. Every value also carries a visual citation linking it to its exact position on the page, and low-confidence fields route to a review queue rather than into your database. In production, L'Oréal reports 99% extraction accuracy and a 60% reduction in processing time across 1,500+ monthly invoices.
Evaluation closes the loop. Every anyformat Extract and Classify workflow has a Health tab where you build a dataset with verified ground truth, run an evaluation that re-extracts every in-scope document against a chosen workflow version, and score it field by field. Runs are numbered and immutable, and Health Overview places the dataset benchmark next to live production accuracy and reports the gap (Evals). Reducto ships a comparable Studio Evaluations feature, but it's gated to its Growth (paid) tier.
Workflow versions matter here too. anyformat's public API refuses to delete or silently overwrite an existing workflow version: every change mints a new one instead. Reducto lets you pin a model version as well, but on a 4-week deprecation clock: a pinned version eventually goes from default to deprecated to removed, so pinning it today still means migrating later.
Long tables and complex layouts
Reducto excels here. RD-TableBench specifically targets complex table extraction, and r-1 handles dense tables, merged cells, and multi-column layouts well.
anyformat's multi-stage pipeline also handles table complexity natively: merged cells, multi-page spans, structural breaks, and nested structures. On tables growing to 50 pages and roughly 2,400 rows it held ~99% row recovery (anyformat benchmark, 2026).
On raw table parsing quality both platforms are strong. The difference is what happens after parsing: validation, calibrated confidence, review routing, and audit through visual citations.
Reducto summarizes figures by default, returning a description of charts and diagrams alongside the parsed text. anyformat detects figures, classifies them in context, and produces structured descriptions. This is parity, and neither product should be chosen on it.
For European buyers, yes, and the decisive gaps are jurisdictional: Reducto is a US company that does not publicly list ISO 27001, where anyformat is EU-native and certified. On product, anyformat adds a human review queue where corrections flow back into the run, calibrated confidence with visual citations for audit, schema management non-technical teams can operate, and an evaluation suite that measures a workflow version against ground truth before it ships.
One honest caveat: if you only need to parse pages and nothing else, Reducto's r-1 is hard to beat on list price today. The comparison changes once the result has to be trustworthy at scale: reviewed by a person when confidence is low, evaluated against your own ground truth before a workflow change ships, and compliant with EU data-residency requirements by default. That's the layer a parsing API alone doesn't give you.
When to choose Reducto
Reducto fits when your team wants a code-first parsing primitive with strong open table benchmarks, when US jurisdiction and SOC 2 Type II satisfy your compliance framework, when list price on parsing alone is the deciding factor, and when your pipeline does not need a supervised review step over uncertain fields.
Choose anyformat when EU jurisdiction and ISO 27001 are hard requirements, when you need calibrated confidence and visual citations to justify straight-through processing to an auditor, when ops teams rather than developers should own the schemas, or when you want workflow changes measured against a ground-truth dataset before they reach production.
anyformat is the agentic document intelligence platform built for European enterprises. ISO 27001 certified, GDPR-compliant, with zero-retention processing and on-premise deployment. Get started at anyformat.ai
