Benchmark Almanac

AncientDoc

Gemini2.5-Pro (snapshot unspecified) 6.37 tops its leaderboard Qwen-VL-Max 5.77 leads the models we track
Not in model cards

Models transcribe historical Chinese pages, translate passages into modern Chinese, and answer questions about meaning, knowledge, and literary style.

The project site scores 17 models on page transcription. GJ.cool independently checked 2,580 reference texts and documented omissions and reading order errors.

Released
August 2025
Kind
Documents, charts and text in images
Level
University
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
Chinese
In and out
Text, Documents → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Last model added
April 2025
Scores refreshed
November 20th, 2026
26
Entries on its own leaderboard, 18 models we track

Gemini2.5-Pro (snapshot unspecified) tops the leaderboard with 6.37 Qwen-VL-Max leads with 5.77 among models released by April 2025

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 Qwen-VL-Max, at 5.77. This view shows only models we track, each at its best.

The newest model with a score here came out in April 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Gemini2.5-Pro (snapshot unspecified) 6.37
  2. 2 Qwen3-VL-8B 6.16
  3. 3 Doubao-V2 5.79
  4. 4 Qwen-VL-Max 5.77
  5. 5 Qwen2.5-VL-72B 5.72
  6. 6 Qwen2.5-VL-32B 5.70
  7. 7 Qwen3-VL-4B 5.52
  8. 8 Qwen2.5-VL-7B 5.02
  9. 9 GPT-4o (snapshot unspecified) 4.47
  10. 10 Qwen2-VL-7B-Instruct 4.02
  11. 11 Qwen2.5-Omni-7B 3.97
  12. 12 InternVL3-78B 3.95
  13. 13 InternVL3-38B 3.68
  14. 14 Qwen2.5-VL-3B 3.68
  15. 15 InternVL3-8B 3.63
  16. 16 InternVL2.5-8B 3.60
  17. 17 Qwen2.5-Omni-3B 3.51
  18. 18 InternVL2.5-4B 3.22
  19. 19 Qwen2-VL-2B-Instruct 3.03
  20. 20 LLaVA-OneVision 72B 2.80
  21. 21 InternVL3-2B 2.67
  22. 22 InternVL2.5-2B 2.07
  23. 23 InternVL2.5-1B 1.66
  24. 24 InternVL3-1B 1.61
  25. 25 LLaVA1.5-7B 0.67
  26. 26 LLaVA-Onevision-5B 0.38

Source: Official AncientDoc leaderboard

#EntryScoreSettingDateSource
1 Gemini2.5-Pro (snapshot unspecified) 6.37 Official leaderboard
2 Qwen3-VL-8B 6.16 Official leaderboard
3 Doubao-V2 5.79 Official leaderboard
4 Qwen-VL-Max 5.77 Jan 2024 Official leaderboard
5 Qwen2.5-VL-72B 5.72 Jan 2025 Official leaderboard
6 Qwen2.5-VL-32B 5.70 Mar 2025 Official leaderboard
7 Qwen3-VL-4B 5.52 Official leaderboard
8 Qwen2.5-VL-7B 5.02 Jan 2025 Official leaderboard
9 GPT-4o (snapshot unspecified) 4.47 Official leaderboard
10 Qwen2-VL-7B-Instruct 4.02 Aug 2024 Official leaderboard
11 Qwen2.5-Omni-7B 3.97 Mar 2025 Official leaderboard
12 InternVL3-78B 3.95 Apr 2025 Official leaderboard
13 InternVL3-38B 3.68 Apr 2025 Official leaderboard
14 Qwen2.5-VL-3B 3.68 Jan 2025 Official leaderboard
15 InternVL3-8B 3.63 Apr 2025 Official leaderboard
16 InternVL2.5-8B 3.60 Dec 2024 Official leaderboard
17 Qwen2.5-Omni-3B 3.51 Official leaderboard
18 InternVL2.5-4B 3.22 Dec 2024 Official leaderboard
19 Qwen2-VL-2B-Instruct 3.03 Aug 2024 Official leaderboard
20 LLaVA-OneVision 72B 2.80 Aug 2024 Official leaderboard
21 InternVL3-2B 2.67 Apr 2025 Official leaderboard
22 InternVL2.5-2B 2.07 Dec 2024 Official leaderboard
23 InternVL2.5-1B 1.66 Dec 2024 Official leaderboard
24 InternVL3-1B 1.61 Apr 2025 Official leaderboard
25 LLaVA1.5-7B 0.67 Official leaderboard
26 LLaVA-Onevision-5B 0.38 Official leaderboard
#ModelScoreSettingDateSource
1 Qwen-VL-Max 5.77 Jan 2024 Official leaderboard
2 Qwen2.5-VL-72B 5.72 Jan 2025 Official leaderboard
3 Qwen2.5-VL-32B 5.70 Mar 2025 Official leaderboard
4 Qwen2.5-VL-7B 5.02 Jan 2025 Official leaderboard
5 Qwen2-VL-7B-Instruct 4.02 Aug 2024 Official leaderboard
6 Qwen2.5-Omni-7B 3.97 Mar 2025 Official leaderboard
7 InternVL3-78B 3.95 Apr 2025 Official leaderboard
8 InternVL3-38B 3.68 Apr 2025 Official leaderboard
9 Qwen2.5-VL-3B 3.68 Jan 2025 Official leaderboard
10 InternVL3-8B 3.63 Apr 2025 Official leaderboard
11 InternVL2.5-8B 3.60 Dec 2024 Official leaderboard
12 InternVL2.5-4B 3.22 Dec 2024 Official leaderboard
13 Qwen2-VL-2B-Instruct 3.03 Aug 2024 Official leaderboard
14 LLaVA-OneVision 72B 2.80 Aug 2024 Official leaderboard
15 InternVL3-2B 2.67 Apr 2025 Official leaderboard
16 InternVL2.5-2B 2.07 Dec 2024 Official leaderboard
17 InternVL2.5-1B 1.66 Dec 2024 Official leaderboard
18 InternVL3-1B 1.61 Apr 2025 Official leaderboard

Rankings by part

We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Page-level OCR, Vernacular Translation and Reasoning-based QA. Qwen-VL-Max leads 2 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Page-level OCR

18 models we track have a score here.

  1. 1 Qwen2.5-VL-7B ↑3 6.37
  2. 2 Qwen2.5-VL-72B 5.73
  3. 3 Qwen-VL-Max ↓2 5.70
  4. 4 Qwen2.5-VL-32B ↓1 5.67
  5. 5 Qwen2.5-VL-3B ↑4 5.56
  6. 6 InternVL3-8B ↑4 5.06
  7. 7 Qwen2-VL-7B-Instruct ↓2 4.66
  8. 8 Qwen2.5-Omni-7B ↓2 4.30
  9. 9 InternVL3-78B ↓2 4.04
  10. 10 Qwen2-VL-2B-Instruct ↑3 3.79

Number one over time

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

Other labs
Own leaderboard Not tested since Aug 2025

Source: Official AncientDoc leaderboard

Scores over time

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
1.00 2.00 3.00 4.00 5.00 6.00 2024 2025 1.00 2.00 3.00 4.00 5.00 6.00 2024 2025

Source: Official AncientDoc 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 Newest score check. The newest came 19 months ago, in April 2025, for InternVL3-1B. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (26 entries).
  • Pass Models scored 18 models we track, plus 8 other entries on its leaderboard.
  • Warning Newest score The newest came 19 months ago, in April 2025, for InternVL3-1B.
  • Note Frontier models None of its models are on the capabilities index.
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on bytedance.github.io: 18 rows for models we track and 8 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