Benchmark Almanac

OBI-Bench

GPT-4o Aug 2024 0.61 leads since October 2024
Not in model cards

Models locate inscriptions, rejoin bone fragments, classify characters, retrieve matches, and explain glyphs. The authors publish separate scores for each task.

The authors publish task results for 23 vision language models across the suite. The benchmark draws 5,523 images from 11 sources.

Released
October 2024
Kind
Documents, charts and text in images
Level
Expert
Scored
Checked automatically
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
Chinese
In and out
Text, Images → Text, Images
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-4o Aug 2024 #167 on the capability index
Last model added
August 2024
Scores refreshed
November 20th, 2026
23
Entries in its paper, 15 models we track

GPT-4o leads with 0.61 among models released by August 2024 GPT-4o leads with 0.61 among models released by August 2024

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. The paper gives no overall score, so the headline is cosine similarity on the O2BR recognition question “What”. 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. This view shows only models we track, each at its best.

The newest model with a score here came out in August 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-4o Aug 2024 0.61
  2. 2 Qwen-VL-Max (ver. 0809) 0.61
  3. 3 InternVL2-Llama3-76B 0.58
  4. 4 Gemini 1.5 Pro (snapshot unspecified) 0.57
  5. 5 InternVL2-40B 0.57
  6. 6 MiniCPM-V-2_6 0.56
  7. 7 GPT-4V 0.54
  8. 8 GLM-4V-9B 0.54
  9. 9 CogVLM2-19B-Chat 0.53
  10. 10 GLM-4V 0.53
  11. 11 xGen-MM (Instruct-interleave-4B) 0.52
  12. 12 InternVL2-8B 0.52
  13. 13 DeepSeek-VL (DeepSeek-LLM-7B) 0.51
  14. 14 InternLM-XComposer2-VL-7B 0.51
  15. 15 LLaVA-NeXT-72B 0.48
  16. 16 Moondream2 (ver. 0728) 0.48
  17. 17 llama3-llava-next-8b 0.48
  18. 18 Qwen-VL (Qwen-7B) 0.45
  19. 19 LLaVA-v1.5-13B 0.44
  20. 20 llava-v1.5-7b 0.42
  21. 21 Gemini 1.5 Flash (snapshot unspecified) 0.41
  22. 22 mPLUG-Owl3-7B 0.33
  23. 23 Idefics2-8B 0.32

Source: The OBI-Bench paper

#EntryScoreSettingDateSource
1 GPT-4o Aug 2024 0.61 Aug 2024 Official leaderboard
2 Qwen-VL-Max (ver. 0809) 0.61 Official leaderboard
3 InternVL2-Llama3-76B 0.58 Jul 2024 Official leaderboard
4 Gemini 1.5 Pro (snapshot unspecified) 0.57 Official leaderboard
5 InternVL2-40B 0.57 Jul 2024 Official leaderboard
6 MiniCPM-V-2_6 0.56 Aug 2024 Official leaderboard
7 GPT-4V 0.54 Nov 2023 Official leaderboard
8 GLM-4V-9B 0.54 Jun 2024 Official leaderboard
9 CogVLM2-19B-Chat 0.53 May 2024 Official leaderboard
10 GLM-4V 0.53 Official leaderboard
11 xGen-MM (Instruct-interleave-4B) 0.52 Official leaderboard
12 InternVL2-8B 0.52 Jul 2024 Official leaderboard
13 DeepSeek-VL (DeepSeek-LLM-7B) 0.51 Official leaderboard
14 InternLM-XComposer2-VL-7B 0.51 Jan 2024 Official leaderboard
15 LLaVA-NeXT-72B 0.48 May 2024 Official leaderboard
16 Moondream2 (ver. 0728) 0.48 Official leaderboard
17 llama3-llava-next-8b 0.48 Apr 2024 Official leaderboard
18 Qwen-VL (Qwen-7B) 0.45 Official leaderboard
19 LLaVA-v1.5-13B 0.44 Oct 2023 Official leaderboard
20 llava-v1.5-7b 0.42 Oct 2023 Official leaderboard
21 Gemini 1.5 Flash (snapshot unspecified) 0.41 Official leaderboard
22 mPLUG-Owl3-7B 0.33 Aug 2024 Official leaderboard
23 Idefics2-8B 0.32 Apr 2024 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-4o Aug 2024 0.61 Aug 2024 Official leaderboard
2 InternVL2-Llama3-76B 0.58 Jul 2024 Official leaderboard
3 InternVL2-40B 0.57 Jul 2024 Official leaderboard
4 MiniCPM-V-2_6 0.56 Aug 2024 Official leaderboard
5 GPT-4V 0.54 Nov 2023 Official leaderboard
6 GLM-4V-9B 0.54 Jun 2024 Official leaderboard
7 CogVLM2-19B-Chat 0.53 May 2024 Official leaderboard
8 InternVL2-8B 0.52 Jul 2024 Official leaderboard
9 InternLM-XComposer2-VL-7B 0.51 Jan 2024 Official leaderboard
10 LLaVA-NeXT-72B 0.48 May 2024 Official leaderboard
11 llama3-llava-next-8b 0.48 Apr 2024 Official leaderboard
12 LLaVA-v1.5-13B 0.44 Oct 2023 Official leaderboard
13 llava-v1.5-7b 0.42 Oct 2023 Official leaderboard
14 mPLUG-Owl3-7B 0.33 Aug 2024 Official leaderboard
15 Idefics2-8B 0.32 Apr 2024 Official leaderboard

Rankings by part

We read 38 parts from the benchmark’s paper in 5 groups: recognition, rejoining, classification, retrieval and deciphering. Each ranks models on one part of the test. GPT-4o leads 27 of them, and 5 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

O2BR: What

15 models we track have a score here.

  1. 1 GPT-4o Aug 2024 0.61
  2. 2 InternVL2-Llama3-76B 0.58
  3. 3 InternVL2-40B 0.57
  4. 4 MiniCPM-V-2_6 0.56
  5. 5 GPT-4V 0.54
  6. 6 GLM-4V-9B 0.54
  7. 7 CogVLM2-19B-Chat 0.53
  8. 8 InternVL2-8B 0.52
  9. 9 InternLM-XComposer2-VL-7B 0.51
  10. 10 LLaVA-NeXT-72B 0.48

Number one over time

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

OpenAI
Benchmark paper Not tested since Oct 2024
GPT-4o 0.61

Source: The OBI-Bench paper

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
In the benchmark’s paper
0.40 0.50 0.60 0.70 2024 0.40 0.50 0.60 0.70 2024

Source: Official OBI-Bench 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 2 years ago, in August 2024, for mPLUG-Owl3-7B. 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 benchmark’s paper (23 entries).
  • Pass Models scored 15 models we track, plus 8 other entries on its leaderboard.
  • Warning Newest score The newest came 2 years ago, in August 2024, for mPLUG-Owl3-7B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-4o (#167).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 15 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