Benchmark Almanac

OmniParseBench

Gemini 4 Argon 95.8 leads since October 2026
Not in model cards

Document parsers convert single pages to text, and 16,288 binary unit tests on 2,937 pages check the output. The headline joins text, table and layout subscores.

The results table also lists cost per 1,000 pages and refusals. Code is Apache 2.0 and page data keeps each source's license. Datalab includes its own parsers.

Made by
Datalab
Released
October 2026
Kind
Documents, charts and text in images
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Strongest model tested
Gemini 4 Carbon #2 on the capability index
Last model added
October 2026
Scores refreshed
November 12th, 2026
34
Entries on its own leaderboard, 9 models we track

Gemini 4 Argon leads with 95.8

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

34 entries. Scroll sideways for the rest.

  1. 95.8 Gemini 4 Argon
  2. 95.6 Gemini 4 Carbon
  3. 94.1 Datalab accurate
  4. 92.9 Gemini 3.8 Flash
  5. 92.6 Datalab balanced
  6. 91.5 GPT-6.1 Sol
  7. 90.8 Reducto
  8. 90.4 LlamaParse
  9. 90.1 chandra-ocr-2 (Datalab)
  10. 90.0 Claude Opus 5.5
  11. 89.5 Mistral OCR
  12. 88.7 Claude Sonnet 5.5
  13. 87.9 dots.mocr
  14. 87.8 GPT-5.6 Sol
  15. 87.1 MinerU2.5-Pro
  16. 87.0 Infinity-Parser2-Pro
  17. 85.9 Surya OCR 2.5 (Datalab)
  18. 85.3 OXR-1.0
  19. 84.2 OvisOCR3
  20. 83.3 Extend
  21. 83.2 MiniCPM-OCR-1.0
  22. 83.1 Surya OCR 2 (Datalab)
  23. 83.0 Azure
  24. 82.9 LightOnOCR-3-4B
  25. 81.6 LightOnOCR-3-0.8B
  26. 81.5 GLM-OCR
  27. 81.2 Unlimited-OCR 2
  28. 80.0 Nemotron 3.5 Parse
  29. 79.3 Unlimited-OCR
  30. 78.7 PaddleOCR-VL-1.6
  31. 78.4 Infinity-Parser2-Flash
  32. 75.2 DeepSeek-OCR 2
  33. 69.5 Indra OCR (Datalab)
  34. 64.6 HunyuanOCR-1.5
  1. 1 Gemini 4 Argon 95.8
  2. 2 Gemini 4 Carbon 95.6
  3. 3 Datalab accurate 94.1
  4. 4 Gemini 3.8 Flash 92.9
  5. 5 Datalab balanced 92.6
  6. 6 GPT-6.1 Sol 91.5
  7. 7 Reducto 90.8
  8. 8 LlamaParse 90.4
  9. 9 chandra-ocr-2 (Datalab) 90.1
  10. 10 Claude Opus 5.5 90.0
  11. 11 Mistral OCR 89.5
  12. 12 Claude Sonnet 5.5 88.7
  13. 13 dots.mocr 87.9
  14. 14 GPT-5.6 Sol 87.8
  15. 15 MinerU2.5-Pro 87.1
  16. 16 Infinity-Parser2-Pro 87.0
  17. 17 Surya OCR 2.5 (Datalab) 85.9
  18. 18 OXR-1.0 85.3
  19. 19 OvisOCR3 84.2
  20. 20 Extend 83.3
  21. 21 MiniCPM-OCR-1.0 83.2
  22. 22 Surya OCR 2 (Datalab) 83.1
  23. 23 Azure 83.0
  24. 24 LightOnOCR-3-4B 82.9
  25. 25 LightOnOCR-3-0.8B 81.6
  26. 26 GLM-OCR 81.5
  27. 27 Unlimited-OCR 2 81.2
  28. 28 Nemotron 3.5 Parse 80.0
  29. 29 Unlimited-OCR 79.3
  30. 30 PaddleOCR-VL-1.6 78.7
  31. 31 Infinity-Parser2-Flash 78.4
  32. 32 DeepSeek-OCR 2 75.2
  33. 33 Indra OCR (Datalab) 69.5
  34. 34 HunyuanOCR-1.5 64.6

Source: Official OmniParseBench leaderboard

#EntryScoreSettingDateSource
1 Gemini 4 Argon 95.8 Sep 2026 Official leaderboard
2 Gemini 4 Carbon 95.6 Oct 2026 Official leaderboard
3 Datalab accurate 94.1 Official leaderboard
4 Gemini 3.8 Flash 92.9 Sep 2026 Official leaderboard
5 Datalab balanced 92.6 Official leaderboard
6 GPT-6.1 Sol 91.5 Sep 2026 Official leaderboard
7 Reducto 90.8 Official leaderboard
8 LlamaParse 90.4 Official leaderboard
9 chandra-ocr-2 (Datalab) 90.1 Official leaderboard
10 Claude Opus 5.5 90.0 Sep 2026 Official leaderboard
11 Mistral OCR 89.5 Official leaderboard
12 Claude Sonnet 5.5 88.7 Sep 2026 Official leaderboard
13 dots.mocr 87.9 Official leaderboard
14 GPT-5.6 Sol 87.8 Jul 2026 Official leaderboard
15 MinerU2.5-Pro 87.1 Official leaderboard
16 Infinity-Parser2-Pro 87.0 Official leaderboard
17 Surya OCR 2.5 (Datalab) 85.9 Official leaderboard
18 OXR-1.0 85.3 Official leaderboard
19 OvisOCR3 84.2 Official leaderboard
20 Extend 83.3 Official leaderboard
21 MiniCPM-OCR-1.0 83.2 Official leaderboard
22 Surya OCR 2 (Datalab) 83.1 Official leaderboard
23 Azure 83.0 Official leaderboard
24 LightOnOCR-3-4B 82.9 Official leaderboard
25 LightOnOCR-3-0.8B 81.6 Official leaderboard
26 GLM-OCR 81.5 Jan 2026 Official leaderboard
27 Unlimited-OCR 2 81.2 Official leaderboard
28 Nemotron 3.5 Parse 80.0 Official leaderboard
29 Unlimited-OCR 79.3 Official leaderboard
30 PaddleOCR-VL-1.6 78.7 Official leaderboard
31 Infinity-Parser2-Flash 78.4 Official leaderboard
32 DeepSeek-OCR 2 75.2 Jan 2026 Official leaderboard
33 Indra OCR (Datalab) 69.5 Official leaderboard
34 HunyuanOCR-1.5 64.6 Official leaderboard

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Google DeepMind
Own leaderboard 1 leader
Gemini 4 Argon 95.8

Source: Official OmniParseBench leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
75.0 80.0 85.0 90.0 95.0 100.0 2026 2027 75.0 80.0 85.0 90.0 95.0 100.0 2026 2027

Source: Official OmniParseBench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Vision

All 281 in Vision
  • MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. GPT-5.1 85.4%
  • MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
  • CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
  • MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy. Qwen 3.8 Max 97.7%
  • MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy. Qwen 3.8 Max 85.9%
  • ChartQA In use Models answer human-written and generated questions about chart images that need visual and arithmetic reasoning. Scored by relaxed accuracy with a 5 percent numeric tolerance. Command A Vision 90.9%