Benchmark Almanac

TRUE (Text Recognition and Understanding)

Gemini 2.5 Pro Mar 2025 81.6 leads since March 2025
Not in model cards

Models recognize text in images and answer questions about its content.

The publisher reports model results.

Made by
BAAI FlagEval
Kind
Documents, charts and text in images
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Images, Documents → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
In views
Vision
Strongest model tested
Gemini 2.5 Pro Mar 2025 #104 on the capability index
Last model added
March 2025
Scores refreshed
November 20th, 2026
23
Models scored on its own leaderboard

Gemini 2.5 Pro leads with 81.6 among models released by March 2025

Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.

The newest model with a score here came out in March 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 Gemini 2.5 Pro Mar 2025 81.6
  2. 2 Gemini 2.0 Flash Dec 2024 75.0
  3. 3 Gemini 2.0 Pro 75.0
  4. 4 GPT-4o Nov 2024 66.9
  5. 5 Qwen2.5-VL-72B 64.9
  6. 6 InternVL2_5-78B 60.9
  7. 7 Qwen2.5-VL-32B 59.9
  8. 8 Pixtral Large 57.1
  9. 9 Claude 3.7 Sonnet 51.7
  10. 10 Mistral Small 3.1 51.2
  11. 11 Qwen2.5-VL-7B 49.4
  12. 12 Molmo-72B 47.9
  13. 13 Llama 3.2 90B 41.5
  14. 14 Pixtral 12B 41.0
  15. 15 MiniCPM-o-2_6 40.7
  16. 16 InternVL2.5-8B 38.0
  17. 17 Molmo-7B-D 32.8
  18. 18 Qwen2.5-VL-3B 32.7
  19. 19 LLaVA-OneVision 72B 31.8
  20. 20 LLaVA-OneVision 7B 20.4
  21. 21 Idefics3-8B-Llama3 16.6
  22. 22 Llama 3.2 11B 15.9
  23. 23 Phi-4-multimodal 15.4

Source: Official TRUE (Text Recognition and Understanding) leaderboard

#ModelScoreSettingDateSource
1 Gemini 2.5 Pro Mar 2025 81.6 Mar 2025 Official leaderboard
2 Gemini 2.0 Flash Dec 2024 75.0 Dec 2024 Official leaderboard
3 Gemini 2.0 Pro 75.0 Feb 2025 Official leaderboard
4 GPT-4o Nov 2024 66.9 Nov 2024 Official leaderboard
5 Qwen2.5-VL-72B 64.9 Jan 2025 Official leaderboard
6 InternVL2_5-78B 60.9 Dec 2024 Official leaderboard
7 Qwen2.5-VL-32B 59.9 Mar 2025 Official leaderboard
8 Pixtral Large 57.1 Nov 2024 Official leaderboard
9 Claude 3.7 Sonnet 51.7 Feb 2025 Official leaderboard
10 Mistral Small 3.1 51.2 Mar 2025 Official leaderboard
11 Qwen2.5-VL-7B 49.4 Jan 2025 Official leaderboard
12 Molmo-72B 47.9 Sep 2024 Official leaderboard
13 Llama 3.2 90B 41.5 Sep 2024 Official leaderboard
14 Pixtral 12B 41.0 Sep 2024 Official leaderboard
15 MiniCPM-o-2_6 40.7 Jan 2025 Official leaderboard
16 InternVL2.5-8B 38.0 Dec 2024 Official leaderboard
17 Molmo-7B-D 32.8 Sep 2024 Official leaderboard
18 Qwen2.5-VL-3B 32.7 Jan 2025 Official leaderboard
19 LLaVA-OneVision 72B 31.8 Aug 2024 Official leaderboard
20 LLaVA-OneVision 7B 20.4 Aug 2024 Official leaderboard
21 Idefics3-8B-Llama3 16.6 Aug 2024 Official leaderboard
22 Llama 3.2 11B 15.9 Sep 2024 Official leaderboard
23 Phi-4-multimodal 15.4 Feb 2025 Official leaderboard

Rankings by part

We read 7 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as HW, SceneOCR and FakeBrands. Gemini 2.5 Pro leads 3 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.

HW

23 models we track have a score here.

  1. 1 Gemini 2.0 Flash Dec 2024 ↑1 89.5
  2. 2 Gemini 2.5 Pro Mar 2025 ↓1 88.2
  3. 3 Qwen2.5-VL-72B ↑2 76.3
  4. 4 Gemini 2.0 Pro ↓1 73.7
  5. 5 Mistral Small 3.1 ↑5 73.7
  6. 6 GPT-4o Nov 2024 ↓2 60.5
  7. 7 Claude 3.7 Sonnet ↑2 60.5
  8. 8 Qwen2.5-VL-3B ↑10 57.9
  9. 9 Qwen2.5-VL-32B ↓2 57.9
  10. 10 MiniCPM-o-2_6 ↑5 55.3

9 models have held number one on the official leaderboard since 2023

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

Hugging Face Other labs Mistral AI Meta Ai2 OpenAI Google DeepMind
Own leaderboard Not tested since Mar 2025

Source: Official TRUE (Text Recognition and Understanding) 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
0.00 20.0 40.0 60.0 80.0 100.0 2025 0.00 20.0 40.0 60.0 80.0 100.0 2025

Source: Official TRUE (Text Recognition and Understanding) 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 20 months ago, in March 2025, for Gemini 2.5 Pro. 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 (23 entries).
  • Pass Models scored 23 models we track.
  • Warning Newest score The newest came 20 months ago, in March 2025, for Gemini 2.5 Pro.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 2.5 Pro (#104).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on flageval-baai.github.io: 23 rows for models we track 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