Marker is an open-source document converter from Datalab that turns PDFs, images, Office files and EPUB into Markdown, JSON, HTML or chunks for RAG. The code is Apache 2.0, the model weights use a modified OpenRAIL-M license that is free for research, personal use and startups under $5M in funding or revenue, and the project has about 40,500 GitHub stars (Marker on GitHub, accessed 15 October 2026).
It is fast and capable, and Datalab, the company behind it, also sells a hosted platform. If you are looking for a Marker alternative, anyformat solves an adjacent problem: schema-validated extraction wrapped in workflow orchestration, human review and a managed service.
anyformat is enterprise-ready agentic document processing: agentic OCR that turns any document into data you can trust, where every value is validated and cites its source. It is built for companies with serious demands on security, scale and audit, so process owners and their IT teams use it alongside the developers who evaluate and integrate it.
You describe the fields once as a schema. Parse reads the document, Extract returns each value with a confidence score and its position on the page, Validate runs your rules, and low-confidence values go to a reviewer in the same workflow, through the API, the Python and TypeScript SDKs or the no-code Studio. Every value carries its evidence, it holds up on hard documents, it is a European company with ISO 27001 and GDPR, and the price covers the whole workflow, not only the extraction. Compared with a converter you operate, extraction, review and validation come built in.
The measure that matters is how many documents go through without anyone touching them, and teams go live in days, not quarters, with SDKs, templates and evals to measure accuracy as they go.
What Marker is, and what Datalab adds
Marker reads a document and returns its content: tables, equations, links, code blocks and images, with headers and footers removed by default. It offers a command line, a Python API, a Streamlit app and a small API server that its own README calls "not a very robust API", and an optional --use_llm mode that sends hard cases to a model such as Gemini, Claude or Ollama. OCR runs through Surya, another Datalab model.
Datalab turns the same technology into a hosted platform: Convert for documents to Markdown, HTML or JSON with word-level bounding boxes; Extract for fields against a schema; Segment for splitting combined files; Eval and Monitor for measuring quality over time; and versioned pipelines that chain these steps (Datalab, accessed 15 October 2026). Named customers include Anthropic, ElevenLabs and Siemens Healthineers.
So the real comparison is with two things: the open-source tool, which you operate, and the hosted platform, which Datalab operates.
Customization and extraction approach
Marker's job is conversion, and its output is a document that a later step turns into fields. Datalab's hosted Extract goes further: it extracts specific fields and returns "citations back to source bounding boxes for auditability" (Datalab documentation, accessed 15 October 2026).
anyformat extracts against a user-defined schema in zero-shot and returns validated JSON, with document-level and table-level fields declared together and returned in one call. Every field carries a confidence score and a visual citation pointing at its exact position on the page.
Confidence: how well it read versus whether the value is right
The pages we read describe citations to source boxes in Datalab's Extract and per-region confidence scores in its Chandra demo. They do not describe a confidence for each extracted field value, so a review process has nothing to route on at the field level unless you build it.
anyformat returns a confidence per extracted field, with the source evidence next to it, so a threshold decides which values flow through and which go to a reviewer in the same workflow. For how the scores are produced and read, see how confidence scoring works.
Benchmark: where each one leads
On the independent ParseBench leaderboard, read on 15 October 2026, anyformat (Standard) scored 80.30 overall, fifth of 147 entries. Datalab's hosted API scored 69.95 in its Accurate mode (44th), 69.51 in Balanced (48th) and 67.79 in Fast (57th). Datalab is the company behind Marker. ParseBench is published by LlamaIndex, which also sells a parsing product that tops its own leaderboard.
The category scores show the split, with Datalab's Accurate mode against anyformat:
Datalab is ahead on tables and on grounding values to the page; anyformat is ahead on charts, formatting and faithfulness. Datalab's own README reports a different benchmark, olmocr-bench, where Marker's balanced mode scores 76.0 and its hosted Chandra 2 scores 85.8 (Datalab, vendor-reported). Benchmarks measure parsing, not extraction against your own schema, which is why a test on your own documents is the real comparison.
In our own hands-on test of PDF-to-markdown tools, Marker read the text PDFs with a correct table and the two columns in order, apart from one dropped heading. On a Mac without a GPU it could not convert a scanned page until an extra binary was installed, and on a 40-page PDF with identical body text in every section it dropped the body of sections 2 to 40. We did not install the binary or find the cause of the second case, so treat both as observations, not verdicts (PDF to markdown for LLMs, 30 September 2026).
Accuracy: reading, structure and values
Accuracy means three different things in document AI. Reading the characters and rebuilding structure (tables, reading order, formulas) is where Marker and Datalab are strongest, and the ParseBench table score shows it. Getting the right value into the right field is extraction, and it is what a downstream system consumes: a conversion score cannot tell you whether the invoice total landed in the total field, or whether row 1,800 of a long table survived. anyformat's long-document study held about 99% row recovery on tables running to 50 pages and roughly 2,400 rows (our benchmark, 2026).
If accuracy decides the choice, test it the way you will run it. Take 50 to 100 of your own documents, extract the same schema with both, and score at field level: exact match for identifiers and amounts, row recall for tables. Then check whether the confidence each system reports predicts which fields are wrong, because that decides how much a person has to review. anyformat's evals suite runs that kind of comparison on your own documents before a workflow goes live.
Run both on your own documents. Try anyformat free, with a one-time grant of 50,000 credits on Pay As You Go, or book a demo.
Marker's code is free, but its weights are not free for everyone: above $5M in funding or revenue, commercial use needs a license from Datalab (Datalab pricing, accessed 15 October 2026). Running it also costs infrastructure: balanced mode wants a GPU, and the CPU mode needs a llama.cpp server binary.
Datalab's hosted API is priced per 1,000 pages: Convert is $4 in fast and balanced mode and $10 in accurate mode, and Extraction runs from $6 to $20 depending on the tier, before extra fees. The Free plan includes a monthly allowance of $10 to $20, the Team plan is $400 a month, and EU processing adds 25% (Datalab pricing, accessed 15 October 2026).
anyformat bills in credits, charged per page and per operator: 10 credits cost €0.01, so Parse is 25 credits per page (€0.025, published as $25 per 1,000 pages), Extract 35 credits per page, and Validate 5 credits per AI rule, once per extraction (anyformat pricing, 2026). Parse plus Extract is 60 credits, or €0.06, per page.
On conversion, Datalab is cheaper per page, and that is a real difference. The two numbers are not the same product: anyformat's price includes the review queue, the validation and the orchestration that you would otherwise build and run yourself.
Compliance and European sovereignty
anyformat is ISO 27001 certified, GDPR-compliant and an EU-headquartered company with an EU legal entity, with compliance included at every plan and no surcharge.
Marker the open-source tool carries no certifications of its own: compliance depends on where and how you run it. Datalab's hosted platform lists SOC 2 Type II, and its README describes zero data retention by default and custom BAAs; for ISO 27001, GDPR terms and hosting regions, check Datalab's trust and data-processing documents directly, as this page does not restate them.
Workflow builder, review and validation
Marker converts documents. Datalab adds versioned pipelines that chain its processors, plus Segment, Eval and Monitor. The pages we read describe no no-code workflow builder and no human review queue: the step where a person checks a doubtful value is yours to build.
anyformat includes Studio, a no-code workflow canvas: branching, conditions, splitting multi-document packets, routing, extraction operators, validation rules, cross-referencing against internal data sources, and human-in-the-loop review placed wherever the process needs it. Low-confidence fields route to a reviewer, who sees the value next to its source on the page and can correct it. Operations teams change the flow without a code deploy.
Version stability
With an open-source tool you pin the package version and the downloaded models yourself, and with a hosted API you follow the vendor's model updates; Datalab sells Monitor precisely to catch regressions across them.
anyformat workflow versions are pinned and immutable: a published version_id does not change underneath you, which matters when a regulated process has to reprocess a document identically years later.
Yes, if you want extraction to run as a process rather than a call: schema-validated output with per-field confidence, a review queue, validation rules, no-code orchestration and version-pinned workflows, without operating models and GPUs yourself. If you want Markdown out of documents at the lowest price per page, or open weights you can run yourself, it depends on the job: Marker and Datalab do that well, and anyformat is not a cheaper way to do it. And if you like Marker's conversion, the two can sit together, because the layer that makes the output production-safe is the part anyformat adds.
When to choose Marker
Fast, low-cost conversion into Markdown or JSON, with open-source code you can run yourself while your company is under the weights license thresholds, or Datalab's hosted API when you want conversion and extraction by the page at a low rate.
When extraction has to run as a process rather than a call: per-field confidence with evidence, validation rules, human review, no-code orchestration and pinned workflow versions, as a managed service. Get started at anyformat.ai.
See also the other comparisons: anyformat vs Docling, anyformat vs LlamaParse and anyformat vs Mistral OCR, or browse all comparisons.
