Skip to content

Benchmark

TextVQA

OCR benchmarks measure how well models read text in images and documents. Scores are percentages.

Models measured
26
Scores recorded
26
Sources
2
IQ weight
1.09

Score distribution

66.1%
Mean
72%
Median
21.3%
Std dev
36.9%
10th percentile
84.4%
90th percentile
8.4%
Min
8.4%85.5%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.09 for chat models, measured over 25 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Qwen2-VL-72B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
85.5%
2Qwen2.5-VL 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim84.9%
3Qwen2.5-Omni 7B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim84.4%
4DeepSeek VL2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim84.2%
5InternVL2 26B· OpenVLM Leaderboard data82.5%
6internvl3-8b· OpenVLM Leaderboard data82.1%
7Nova Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.5%
8Nova Lite· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim80.2%
9MiniCPM-V 2.6· OpenVLM Leaderboard data78.3%
10Phi 4 Multimodal Instruct· OpenVLM Leaderboard data
Values from 2 sources
76.9%
11Pixtral 12B· OpenVLM Leaderboard data75.7%
12InternVL2 4B· OpenVLM Leaderboard data74.7%
13Phi-3.5 Vision Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim72%
14GPT 4.1· OpenVLM Leaderboard data71%
15Molmo 7B-D· OpenVLM Leaderboard data70.3%
16Gemma 3 12B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
67.7%
17Gemma 3 27B IT· OpenVLM Leaderboard data
Values from 2 sources
66.7%
18GPT-4.1 mini· OpenVLM Leaderboard data64.9%
19Gemini 1.5 Flash· OpenVLM Leaderboard data61.6%
20GPT-4.1 nano· OpenVLM Leaderboard data60.9%
21Gemma 3 4B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim57.8%
22Gemma 3 4B· OpenVLM Leaderboard data57.7%
23gemini-1.0-pro· OpenVLM Leaderboard data36.9%
24kosmos-2· OpenVLM Leaderboard data14.7%
25Qwen-VL Max· OpenVLM Leaderboard data8.6%
26Qwen-VL Plus· OpenVLM Leaderboard data8.4%

Data credits: Data by LLM Stats · OpenVLM Leaderboard (OpenCompass / VLMEvalKit), Apache-2.0 (licence)