Benchmark Almanac

MTHv2

TongGuOCR 97.9% tops its leaderboard Kimi K2.6 86.4% leads the models we track
Not in model cards

Models read Chinese Buddhist texts from scanned historical pages. The authors annotate characters, line transcriptions, reading order, and boundary lines.

TongGuOCR reports 11 general vision language models on the original 800 test pages. Some historical baselines receive cropped lines.

Released
June 2020
Kind
Documents, charts and text in images
Level
University
Scored
Checked automatically
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
Chinese
In and out
Documents → Text
Interaction
Single turn
Items from
Real work and use
Run by
Anyone
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
June 2026
Scores refreshed
November 20th, 2026
31
Entries on TongGuOCR’s leaderboard, 11 models we track

TongGuOCR tops the leaderboard with 97.9% Kimi K2.6 leads with 86.4% among models released by June 2026

Best score per entry from TongGuOCR’s 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 Kimi K2.6, at 86.4%. This view shows only models we track, each at its best.

The newest model with a score here came out in June 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 TongGuOCR Page recognition 97.9%
  2. 2 ZCTRN Text-line recognition 97.4%
  3. 3 CRNN Text-line recognition 96.9%
  4. 4 Liu et al. Page recognition 96.3%
  5. 5 Ma et al. Page recognition 95.5%
  6. 6 HanDoc-OrderOCR Page recognition 95.3%
  7. 7 RAN Text-line recognition 91.6%
  8. 8 Kimi K2.6 Page recognition 86.4%
  9. 9 Qwen3.7-Plus Page recognition 81.2%
  10. 10 Gemini 3.1 Pro Page recognition 78.1%
  11. 11 HunyuanOCR Page recognition 77.8%
  12. 12 MinerU2.5-Pro Page recognition 76%
  13. 13 PaddleOCR-VL-1.6 Page recognition 72.3%
  14. 14 GPT-5.5 Page recognition 70.2%
  15. 15 GLM-OCR Page recognition 69.5%
  16. 16 MiniMax-M3 Page recognition 67.5%
  17. 17 Qianfan-OCR Page recognition 62.8%
  18. 18 Qwen3.5-9B Page recognition 62.2%
  19. 19 MinerU2.5 Page recognition 48.7%
  20. 20 PaddleOCR-VL-1.5 Page recognition 47.4%
  21. 21 dots.ocr Page recognition -3.27%
  22. 22 Qwen3-VL-8B Page recognition -29.34%
  23. 23 InternVL3.5-8B Page recognition -38.99%
  24. 24 Qwen3.5-0.8B Page recognition -69.63%
  25. 25 InternVL3.5-2B Page recognition -126.62%
  26. 26 MonkeyOCR-3B Page recognition -184.49%
  27. 27 FireRed-OCR Page recognition -198.23%
  28. 28 Qwen3-VL-2B Page recognition -376.72%
  29. 29 MonkeyOCR-1.2B Page recognition -432.67%
  30. 30 DeepSeek-OCR Page recognition -494.24%
  31. 31 DeepSeek-OCR 2 Page recognition -540.58%

Source: TongGuOCR’s leaderboard

#EntryScoreSettingDateSource
1 TongGuOCR 97.9% Page recognition Official leaderboard
2 ZCTRN 97.4% Text-line recognition Official leaderboard
3 CRNN 96.9% Text-line recognition Official leaderboard
4 Liu et al. 96.3% Page recognition Official leaderboard
5 Ma et al. 95.5% Page recognition Official leaderboard
6 HanDoc-OrderOCR 95.3% Page recognition Official leaderboard
7 RAN 91.6% Text-line recognition Official leaderboard
8 Kimi K2.6 86.4% Page recognition Apr 2026 Official leaderboard
9 Qwen3.7-Plus 81.2% Page recognition Jun 2026 Official leaderboard
10 Gemini 3.1 Pro 78.1% Page recognition Feb 2026 Official leaderboard
11 HunyuanOCR 77.8% Page recognition Official leaderboard
12 MinerU2.5-Pro 76% Page recognition Official leaderboard
13 PaddleOCR-VL-1.6 72.3% Page recognition Official leaderboard
14 GPT-5.5 70.2% Page recognition Apr 2026 Official leaderboard
15 GLM-OCR 69.5% Page recognition Jan 2026 Official leaderboard
16 MiniMax-M3 67.5% Page recognition Jun 2026 Official leaderboard
17 Qianfan-OCR 62.8% Page recognition Official leaderboard
18 Qwen3.5-9B 62.2% Page recognition Feb 2026 Official leaderboard
19 MinerU2.5 48.7% Page recognition Official leaderboard
20 PaddleOCR-VL-1.5 47.4% Page recognition Official leaderboard
21 dots.ocr -3.27% Page recognition Official leaderboard
22 Qwen3-VL-8B -29.34% Page recognition Official leaderboard
23 InternVL3.5-8B -38.99% Page recognition Aug 2025 Official leaderboard
24 Qwen3.5-0.8B -69.63% Page recognition Feb 2026 Official leaderboard
25 InternVL3.5-2B -126.62% Page recognition Official leaderboard
26 MonkeyOCR-3B -184.49% Page recognition Official leaderboard
27 FireRed-OCR -198.23% Page recognition Official leaderboard
28 Qwen3-VL-2B -376.72% Page recognition Official leaderboard
29 MonkeyOCR-1.2B -432.67% Page recognition Official leaderboard
30 DeepSeek-OCR -494.24% Page recognition Oct 2025 Official leaderboard
31 DeepSeek-OCR 2 -540.58% Page recognition Jan 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Kimi K2.6 86.4% Page recognition Apr 2026 Official leaderboard
2 Qwen3.7-Plus 81.2% Page recognition Jun 2026 Official leaderboard
3 Gemini 3.1 Pro 78.1% Page recognition Feb 2026 Official leaderboard
4 GPT-5.5 70.2% Page recognition Apr 2026 Official leaderboard
5 GLM-OCR 69.5% Page recognition Jan 2026 Official leaderboard
6 MiniMax-M3 67.5% Page recognition Jun 2026 Official leaderboard
7 Qwen3.5-9B 62.2% Page recognition Feb 2026 Official leaderboard
8 InternVL3.5-8B -38.99% Page recognition Aug 2025 Official leaderboard
9 Qwen3.5-0.8B -69.63% Page recognition Feb 2026 Official leaderboard
10 DeepSeek-OCR -494.24% Page recognition Oct 2025 Official leaderboard
11 DeepSeek-OCR 2 -540.58% Page recognition Jan 2026 Official leaderboard

4 models have held number one on TongGuOCR’s leaderboard since 2023

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

Shanghai AI Lab Z.ai Google DeepMind Moonshot AI
TongGuOCR leaderboard Passed 85% in Apr 2026, by Kimi K2.6
Kimi K2.6 86.4%

Source: TongGuOCR’s leaderboard

Kimi K2.6 on the official leaderboard passed 85% in April 2026

That was 70 months after its release. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On TongGuOCR’s leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Kimi K2.6 Gemini 3.1 Pro GLM-OCR InternVL3.5-8B 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Kimi K2.6 Gemini 3.1 Pro

Source: TongGuOCR’s 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

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the benchmark’s paper (31 entries).
  • Pass Models scored 11 models we track, plus 20 other entries on its leaderboard.
  • Pass Newest score The newest is from June 2026, for Qwen3.7-Plus.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is GPT-5.5 (#21).
  • Note Headroom Saturated. The best is 86.4%, past 85% since April 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on TongGuOCR: 11 rows for models we track and 20 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