Hugging Face's logo

Datasets:
datalab-to
/
omni_parse_bench

Tasks:
Image-to-Text
Modalities:
Document
Image
Text
Formats:
parquet
Size:
100K - 1M
Tags:
document-parsing
ocr
benchmark
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
7
omni_parse_bench / README.md
hheiden's picture
hheiden
Add OvisOCR, OvisOCR2 and OvisOCR3 results (official runs, opb-public)
d70182b verified
preview code
|
Raw History Blame Contribute Delete
17.1 kB
metadata 
pretty_name: Omni Parse Bench
license: other
license_name: per-source
license_link: >-
  https://huggingface.co/datasets/datalab-to/omni_parse_bench/blob/main/licenses/README.md
task_categories:
  - image-to-text
tags:
  - document-parsing
  - ocr
  - benchmark
size_categories:
  - 1K<n<10K
configs:
  - config_name: default
    data_files:
      - split: test
        path: manifest.parquet
  - config_name: results
    data_files:
      - split: test
        path: results.parquet

Omni Parse Bench

A document-parsing benchmark with 16,288 tests on 2,937 pages.

Results

Every system on our dataset. Scoring, run details and charts.

parser weights headline USD / 1,000 pages text tables layout refusals
Datalab accurate closed 94.1 $7.50 92.0 94.2 96.0 —
Gemini 3.8 Flash* closed 92.9 ≈$37.08 91.3 95.2 92.2 8
Datalab balanced closed 92.6 $3.00 90.1 91.6 96.2 —
GPT-6.1 Sol closed 91.5 $30.95 89.8 95.6 89.1 17
Reducto closed 90.8 $10.00 85.4 91.2 95.8 —
LlamaParse closed 90.4 ≈$56.25 88.2 92.6 90.4 —
chandra-ocr-2 open 90.1 — 87.4 87.4 95.4 —
Claude Opus 5.5 closed 90.0 $68.45‡ 93.9 93.4 82.6 40
Mistral OCR closed 89.5 $4.00 82.8 91.9 93.8 —
Claude Sonnet 5.5* closed 88.7 ≈$34.30 93.1 95.4 77.5 7
dots.mocr open 87.9 — 82.5 88.6 92.5 —
GPT-5.6 Sol closed 87.8 ≈$94.49 85.4 91.6 86.5 17
MinerU2.5-Pro open 87.1 — 77.4 92.1 91.7 —
Infinity-Parser2-Pro open 87.0 — 83.3 84.1 93.5 —
Surya OCR 2.5 open 85.9 — 83.8 78.3 95.7 —
OXR-1.0 open 85.3 — 76.4 88.2 91.4 —
OvisOCR3 open 84.2 — 77.5 84.3 90.6 —
Extend closed 83.3 ≈$33.75 64.1 92.5 93.4 —
MiniCPM-OCR-1.0 open 83.2 — 72.8 89.1 87.6 —
Surya OCR 2 open 83.1 — 80.6 74.3 94.3 —
Azure closed 83.0 $5.00 74.6 81.7 92.8 —
LightOnOCR-3-4B open 82.9 — 83.9 70.8 94.0 —
LightOnOCR-3-0.8B open 81.6 — 81.1 72.1 91.7 —
GLM-OCR open 81.5 — 71.0 84.2 89.3 —
Unlimited-OCR 2 open 81.2 — 72.4 79.7 91.6 —
Nemotron 3.5 Parse open 80.0 — 72.4 77.3 90.2 —
Unlimited-OCR open 79.3 — 70.2 76.5 91.1 —
PaddleOCR-VL-1.6 open 78.7 — 61.7 83.7 90.7 —
Infinity-Parser2-Flash open 78.4 — 78.4 66.5 90.4 —
DeepSeek-OCR-2 open 75.2 — 64.2 70.0 91.4 —
Indra OCR open 69.5 — 60.3 58.1 90.1 —
HunyuanOCR-1.5 open 64.6 — 50.5 61.9 81.4 —
Xiaomi-OCR-0 open none — 73.3 90.4 unsupported —
OvisOCR2 open none — 73.7 79.6 unsupported —
OvisOCR open none — 71.8 77.9 unsupported —
olmOCR-2 open none — 71.3 56.6 unsupported —
Nanonets-OCR2-3B open none — 68.2 62.6 unsupported —
Tesseract open none — 17.6 0.0 unsupported —

Costs are USD per 1,000 pages. Opus 5.5 and GPT-6.1 Sol use the measured full-run reported costs. ‡ Opus is a reported subtotal: 12 refused pages returned no usage or cost. Historical prices retain the existing price chart’s 2026-10-08 basis; ≈ marks values read back from that chart rather than remeasured. Datalab prices include the chart’s 25% data-retention discount. — means self-hosted inference cost was not measured; it does not mean free inference.

* Used in creating and verifying at least some tests.

About the benchmark

Choosing a document parser usually starts with a leaderboard: one number per parser. But nobody parses an average document, and the parser that wins on average can be the one that fails on yours. Omni Parse Bench is built to answer the question you actually have: which parser is best on documents like mine? It is:

  • clear: every test is a yes/no question about one page (is this content in the output, in this order, in a table cell under these headings, in a layout block of this kind?), every score is the share of tests passed, and every score comes with the number of tests and documents behind it, so a thin slice says it is thin;
  • auditable: every test records how its answer was established, and the scorer is open code with no model judging the outputs; and
  • explorable: every test is tagged by what its content is (a table cell, handwriting, a rotated page, Thai), so any slice of the benchmark gets its own score.

It has 16,288 tests on 2,937 pages from 2,343 documents over 22 suites, in 92 languages. Omni Parse Bench is developed by Datalab.

Read more in our blog post. Also see the GitHub for how to run the benchmark and use the scorer.

The design

The benchmark's design: headline, families, tests, args and tags

The headline is the mean of three family scores: text, tables and layout. A family is a group of test types:

test type checks family tests
present this text, number or equation appears text 7,275
order text A comes before text B text 1,189
repeat this line appears exactly N times text 35
table_cell this cell sits under these headings and beside these neighbors tables 5,386
layout_kind the block holding this line is a heading, text or table layout 2,403

A test checks one or more args, each of which is the content at one place on the page. Each arg is tagged:

  • structure: table, math, form and multi_column;
  • rendering: handwriting, tiny, rotated and degraded;
  • role: heading, caption, footnote, list, code, figure and header_footer;
  • script: ISO 15924 codes (Latn, Hani, Deva, ...); and
  • language: ISO 639-3 codes (eng, fra, tha, ...).

A test's tags are the union of its args' tags. Tags never decide whether a test passes or what the headline is; they select tests, so any slice of the benchmark can be scored. A test with several tags counts in each, so per-tag scores overlap and don't average to the headline.

A per-tag score (a "hard slice") is the pass rate over every test carrying that tag, with layout tests left out, so a system without layout output isn't penalised on the tag slices. A test on a page the system refused or failed to answer counts as failed. "Not English" is a test with a language tag other than eng. Script tags (Latn, Grek, ...) are not languages and don't count.

Each tag was decided from the page alone: computed where it can be (a script from Unicode, a font size, a rotation we made), from a source where one exists (publisher markup, form widgets), else by two models shown the content boxed on the page, with a third settling their splits by fixed rules; whatever they couldn't settle was decided by a person, who also checked a random sample of 30 tagged pieces of content and found no wrong tag.

Data

Tests, pages and documents per tag

Documents with a test carrying both tags, for every pair of tags

Every language, by its documents and tests

Results details

Refusals count distinct pages with at least one required output refused by the provider, including content filtering and Gemini's RECITATION stop; — means not audited, rather than zero. Gemini has 8 refusals and 12 other failed pages; Sonnet 5.5 has 7 refusals and 3 other failed pages. GPT-5.6 Sol has 17 refusals; Opus 5.5 has 40; GPT-6.1 Sol has 17, with no other failed pages in those three runs.

Opus 5.5 and GPT-6.1 Sol were each run once on 2026-10-27 using omni-parse-bench 1.0.0, the stock OpenRouter adapter and each model's default reasoning. The three earlier LLM rows were recomputed from their published verdicts and archived failure records, without new inference or changes to the tests. Machine-readable LLM run summaries give the exact scores, denominators and failed/refused page IDs.

Open-weight models are served on their own host (a GPU, or CPU only for Indra OCR) from omni-parse-bench's server images, with the defaults each vendor recommends; each adapter's docstring lists where it differs.

The headline is the mean of three family scores: text (present, order and repeat tests), tables (table_cell) and layout (layout_kind). Tesseract, olmOCR-2, Nanonets-OCR2-3B, OvisOCR, OvisOCR2 and Xiaomi-OCR-0 return no layout, so they have no headline.

Every verdict behind these numbers is in results.parquet: one row per parser and test, with the test's page, document, suite, type, derivation and tags, its result (pass, fail, or unsupported when the parser can't produce that output) and why it failed. Browse it as the results subset in the viewer above, or query it; for instance, the share of handwriting tests each parser passes:

select system, round(100 * avg((result = 'pass')::int), 1) as score, count(*) as tests
from 'hf://datasets/datalab-to/omni_parse_bench/results.parquet'
where result in ('pass', 'fail') and list_contains(tags, 'handwriting')
group by system order by score desc;

Headline versus latency

Headline against latency

The new LLM points († diamonds) use the full-page recorded wall time: HTML plus layout when requested, including failed pages and retries, with 32 workers per model. Both medians are 18.5 seconds; p90 is 37.3 seconds for Opus 5.5 and 43.1 seconds for GPT-6.1 Sol. Historical circles retain the earlier median of successful, unretried HTML calls only. These timing definitions differ and are not directly comparable.

Headline against parameters for the open-weight models

Price per 1,000 pages

Model Headline Reported USD / 1,000 pages Refusals
Claude Opus 5.5 90.00 $68.45† 40
GPT-6.1 Sol 91.50 $30.95 17

One full run per model, 2,937 pages each. Prices include all page outputs requested by the benchmark and response-reported upstream model charges under OpenRouter BYOK. † Opus's price is a reported subtotal: 12 refused pages returned no usage or cost. The earlier systems' chart prices are unchanged.

Headline against price per 1,000 pages

Each parser's pass rate on a tag's tests, minus their median

Layout 

manifest.parquet        one row per page
pages/<page_id>.pdf     the page that gets sent: a one-page PDF, or a .png for an image page
gold/<page_id>.json     its tests: {"tests": [...]}
results.parquet         every parser's verdict on every test (Results above)
licenses/               the license, and the credit it requires, of each source
assets/                 the charts on this card
column
page_id unique; the files carry it
document the document the page comes from; scores count documents as well as tests
suite which part of the benchmark the page came from
doc_path pages/<page_id>.pdf (or .png), relative to the dataset root
gt_path gold/<page_id>.json, relative to the dataset root
outputs the output types its tests read (html, blocks): what to predict for the page

A test says what structure of the output it checks (test_type), in which output (output_type), on what content (args), with what else it needs (settings); how its answer was established (derivation); and what its content is (tags):

{
    "id": "p11_PMC12397559.1_t0_head_17_2",
    "test_type": "table_cell",
    "output_type": "html",
    "derivation": "publisher_markup",
    "tags": ["Latn", "eng", "multi_column", "table"],
    "args": {
        "cell": {"text": "32 (40.0)", "max_diffs": 0},
        "headings": [{"text": "No. (%)", "max_diffs": 0}]
    },
    "settings": {}
}

An arg is {"text", "max_diffs"}, {"number"} or {"latex"}, and its kind says how it is compared: text within max_diffs edits, a number digit for digit, LaTeX by what it renders (or by its plain text, when plain text writes it without loss).

Use 

uv pip install 'omni-parse-bench[benchmark]'

How to run the benchmark on this dataset, grade a run, score a page on its own, and use the adapters: the omni-parse-bench README on GitHub. This dataset needs omni-parse-bench 1.0.0 or newer.

How the tests were built

What a test is made of, where each part of it comes from (its derivation says which source gave its answer), and how tags are decided: docs/building-tests.md and docs/tagging.md in the repository.

License

Each source's license, and the credit it requires, is a file in licenses/; each PubMed Central article's credit is a row of licenses/pmc_articles.csv.

The pages are public documents, pages of public datasets, and copies of them we filled in or rebuilt. A dataset's license covers what its makers could license; the documents it collected keep their owners' rights, and are credited here to their source. Pages whose own license forbids redistribution or commercial use (a PubMed Central article under CC BY-NC, for one) are left out. Two sources ask for non-commercial use and are kept, as fair use, with our reasons in their files: the Ohio Department of Transportation drawings (ohio-dot-LICENSE.txt), and the three Copiste d'un jour pages, whose writers consented to non-commercial use (htr-united-LICENSE.txt).

The tests, and what we added to pages (form fills, handwriting), are Datalab's, under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/legalcode). Tests converted from olmOCR-bench and fr-bench-pdf2md keep their licenses (ODC-BY 1.0 and MIT), and the tests on the three Copiste d'un jour pages, which quote Wikipedia, are CC BY-SA 4.0 (htr-united-LICENSE.txt).

If you hold rights in a page and want it removed, open an issue at https://github.com/datalab-to/omni_parse_bench/issues.

The scorer and harness are Apache 2.0.

suites pages license file
olmo_arxiv_math, olmo_long_tiny_text, olmo_multi_column, olmo_old_scans, olmo_old_scans_math, olmo_table_tests 974 olmocr-bench-LICENSE.txt
tables_derived 528 pmc-LICENSE.txt, with each article's credit in pmc_articles.csv
docparsing_images 71 docparsingbench-LICENSE.txt
parsebench_docs 81 parsebench-LICENSE.txt
realdoc_pdfs, realdoc_pages 210 realdoc-bench-LICENSE.txt
multiling_finepdfs 158 finepdfs-LICENSE.txt
rotation_skew 142 the file of the page it is a rotated copy of: pmc-LICENSE.txt, parsebench-LICENSE.txt or realdoc-bench-LICENSE.txt
forms_filled 34 us-federal-forms-LICENSE.txt and cerfa-LICENSE.txt
forms_federal 31 us-federal-forms-LICENSE.txt
multilingual 42 cerfa-LICENSE.txt and nta-LICENSE.txt
frbench_handwriting 29 fr-bench-pdf2md-LICENSE.txt and cerfa-LICENSE.txt
forms_documentcloud 17 documentcloud-LICENSE.txt
large_format, tiny_plans 87 ohio-dot-LICENSE.txt
handwriting_samples 39 loc-catt-papers-LICENSE.txt, nls-marjory-fleming-LICENSE.txt and htr-united-LICENSE.txt
synth_copies 494 the file of the page it is rebuilt from: realdoc-bench-LICENSE.txt, parsebench-LICENSE.txt, docparsingbench-LICENSE.txt, olmocr-bench-LICENSE.txt or finepdfs-LICENSE.txt

Citation 

@misc{omniparsebench2026,
  title        = {Omni Parse Bench: A Document-Parsing Benchmark},
  author       = {{Datalab}},
  year         = {2026},
  howpublished = {\url{https://github.com/datalab-to/omni_parse_bench}},
}

Please also cite the upstream datasets whose pages or tests this benchmark uses: