Benchmark Almanac

Bill of Lading OCR Benchmark

okraPDF (/v1 production) 96.9% tops its leaderboard Gemini 3.1 Flash-Lite 94.9% leads the models we track
Not in model cards

Models transcribe 25 synthetic bills of lading. A fixed Gemini model extracts 21 fields from each transcription for comparison with the known values.

The maker ranks 24 parsers, including 15 vision models. It publishes separate scores for digital documents, office scans, and simulated faxes.

Made by
okraPDF
Released
August 2026
Kind
Documents, charts and text in images
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Measures
Capability
Language
English
In and out
Documents → Text
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
A host
Strongest model tested
Gemini 3.7 Flash #26 on the capability index
Last model added
August 2026
Scores refreshed
November 20th, 2026
24
Entries on its own leaderboard, 13 models we track

okraPDF (/v1 production) tops the leaderboard with 96.9% Gemini 3.1 Flash-Lite leads with 94.9% among models released by August 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 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. The best model we track is Gemini 3.1 Flash-Lite, at 94.9%. This view shows only models we track, each at its best.

The newest model with a score here came out in August 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 okraPDF (/v1 production) 96.9%
  2. 2 Mistral OCR 96.6%
  3. 3 Qwen3-VL 235B 94.9%
  4. 4 Gemini 3.1 Flash-Lite 94.9%
  5. 5 Claude Sonnet 5 94.5%
  6. 6 Qwen3-VL 8B 93.5%
  7. 7 LlamaParse 92.1%
  8. 8 Reducto 90.8%
  9. 9 GLM-4.6V 89.9%
  10. 10 Gemini 3.7 Flash 89.8%
  11. 11 Mistral Small 3 89.3%
  12. 12 Llama 4 Maverick 89.1%
  13. 13 Chandra (Datalab) 88.9%
  14. 14 Kimi K2.5 88.8%
  15. 15 GPT-5 mini 88%
  16. 16 Gemini 2.5 Flash Sep 2025 87.6%
  17. 17 Claude Haiku 4.5 86.4%
  18. 18 GPT-5.4 Mini 85.1%
  19. 19 Nova 2 Lite 84.1%
  20. 20 GPT-4o mini 83.3%
  21. 21 Tesseract 5 79.1%
  22. 22 pdftotext (poppler) 52.4%
  23. 23 PyMuPDF 52.4%
  24. 24 pdfplumber 52.1%

Source: Official Bill of Lading OCR Benchmark leaderboard

#EntryScoreSettingDateSource
1 okraPDF (/v1 production) 96.9% Official leaderboard
2 Mistral OCR 96.6% Official leaderboard
3 Qwen3-VL 235B 94.9% Official leaderboard
4 Gemini 3.1 Flash-Lite 94.9% Mar 2026 Official leaderboard
5 Claude Sonnet 5 94.5% Jun 2026 Official leaderboard
6 Qwen3-VL 8B 93.5% Official leaderboard
7 LlamaParse 92.1% Official leaderboard
8 Reducto 90.8% Official leaderboard
9 GLM-4.6V 89.9% Dec 2025 Official leaderboard
10 Gemini 3.7 Flash 89.8% Aug 2026 Official leaderboard
11 Mistral Small 3 89.3% Jan 2025 Official leaderboard
12 Llama 4 Maverick 89.1% Apr 2025 Official leaderboard
13 Chandra (Datalab) 88.9% Official leaderboard
14 Kimi K2.5 88.8% Jan 2026 Official leaderboard
15 GPT-5 mini 88% Aug 2025 Official leaderboard
16 Gemini 2.5 Flash Sep 2025 87.6% Sep 2025 Official leaderboard
17 Claude Haiku 4.5 86.4% Oct 2025 Official leaderboard
18 GPT-5.4 Mini 85.1% Mar 2026 Official leaderboard
19 Nova 2 Lite 84.1% Dec 2025 Official leaderboard
20 GPT-4o mini 83.3% Jul 2024 Official leaderboard
21 Tesseract 5 79.1% Official leaderboard
22 pdftotext (poppler) 52.4% Official leaderboard
23 PyMuPDF 52.4% Official leaderboard
24 pdfplumber 52.1% Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 3.1 Flash-Lite 94.9% Mar 2026 Official leaderboard
2 Claude Sonnet 5 94.5% Jun 2026 Official leaderboard
3 GLM-4.6V 89.9% Dec 2025 Official leaderboard
4 Gemini 3.7 Flash 89.8% Aug 2026 Official leaderboard
5 Mistral Small 3 89.3% Jan 2025 Official leaderboard
6 Llama 4 Maverick 89.1% Apr 2025 Official leaderboard
7 Kimi K2.5 88.8% Jan 2026 Official leaderboard
8 GPT-5 mini 88% Aug 2025 Official leaderboard
9 Gemini 2.5 Flash Sep 2025 87.6% Sep 2025 Official leaderboard
10 Claude Haiku 4.5 86.4% Oct 2025 Official leaderboard
11 GPT-5.4 Mini 85.1% Mar 2026 Official leaderboard
12 Nova 2 Lite 84.1% Dec 2025 Official leaderboard
13 GPT-4o mini 83.3% Jul 2024 Official leaderboard

Rankings by part

We read 3 parts from the benchmark’s own leaderboard: Digital, Scan and Fax-hard. Each ranks models on one part of the test. GPT-5 mini leads 1 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Digital

13 models we track have a score here.

  1. 1 GPT-5 mini ↑7 100%
  2. 2 Claude Sonnet 5 100%
  3. 3 Gemini 3.7 Flash ↑1 100%
  4. 4 GPT-5.4 Mini ↑7 99.5%
  5. 5 GLM-4.6V ↓2 99%
  6. 6 Gemini 3.1 Flash-Lite ↓5 99%
  7. 7 Mistral Small 3 ↓2 98%
  8. 8 Llama 4 Maverick ↓2 96.5%
  9. 9 Nova 2 Lite ↑3 94%
  10. 10 Claude Haiku 4.5 93.3%

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 Passed 85% in Aug 2026, by Gemini 3.1 Flash-Lite

Source: Official Bill of Lading OCR Benchmark leaderboard

Mistral Small 3 on the official leaderboard passed 85% in January 2025

Models released before it came out already scored that high. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 3.1 Flash-Lite GLM-4.6V Mistral Small 3 GPT-4o mini 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 3.1 Flash-Lite GPT-4o mini

Source: Official Bill of Lading OCR Benchmark leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Vision

All 636 in Vision
  • MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. Qwen 3.6 Plus 86%
  • 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%
  • MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy. Seed2.1 Pro 90.7%
  • 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%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (24 entries).
  • Pass Models scored 13 models we track, plus 11 other entries on its leaderboard.
  • Pass Newest score The newest is from August 2026, for Gemini 3.7 Flash.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is Gemini 3.7 Flash (#26).
  • Note Headroom Saturated. The best is 94.9%, past 85% since January 2025.
Leaderboard import 3 pass
  • Pass Importer We read the board on okrapdf.com: 13 rows for models we track and 11 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check