Skip to content

Benchmark

DocVQA

Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.

Models and systems measured
21
Scores recorded
22
Sources
2
IQ weight
1.87

Score distribution

92.3%
Mean
93.5%
Median
4.4%
Std dev
87.1%
10th percentile
95.6%
90th percentile
75.8%
Min
75.8%96.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.87 for chat models, measured over 19 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1Qwen2.5 VL 72B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim96.4%
2Qwen2.5-VL 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim95.7%
3Claude 3.5 Sonnet· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim95.2%
4Qwen2.5-Omni 7B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim95.2%
5Mistral Small 3.2 24B Instruct2506· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim94.9%
6Llama 4 Maverick· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim94.4%
7Llama 4 Scout· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim94.4%
8Llama 4 Scouttest split· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.94.4%
9Grok 2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim93.6%
10Nova Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim93.5%
11DeepSeek VL2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim93.3%
12Pixtral Large· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim93.3%
13Grok 2 mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim93.2%
14Phi 4 Multimodal Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim93.2%
15GPT 4o· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim92.8%
16Nova Lite· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim92.4%
17Pixtral 12B2409· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim90.7%
18Llama 3.2 11B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim88.4%
19Gemma 3 12B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim87.1%
20Gemma 3 27B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim86.6%
21Gemma 3 4B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim75.8%

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
DeepSeek V4.1 Flash BaseLLM-Judge · 4-shot· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.95.6%

Data credits: Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.