Models
27 models evaluated across 3 document AI benchmarks.
1
Nanonets OCR-3
Nanonets
Overall
85.9
2
Nanonets OCR2+
Nanonets
Overall
81.8
3
GPT-5.4
OpenAI
Overall
81.0
4
Qwen3-VL-Plus
Alibaba
Overall
80.1
5
Qwen3-VL-235B
Alibaba
Overall
79.6
6
Gemini-3-Pro
Overall
79.4
7
Claude Sonnet 4.6
Anthropic
Overall
79.1
8
Claude Opus 4.6
Anthropic
Overall
78.8
9
Gemini-3-Flash
Overall
78.6
10
Gemini 3.1 Pro
Overall
78.5
11
GPT-5.2
OpenAI
Overall
78.0
12
Qwen3.5-9B
Alibaba
Overall
76.7
13
Qwen3.5-4B
Alibaba
Overall
72.5
14
GPT-5-Mini
OpenAI
Overall
71.7
15
Mistral Small 4
Mistral AI
Overall
71.5
16
Claude Haiku 4.5
Anthropic
Overall
70.2
17
Ministral-8B
Mistral AI
Overall
69.5
18
GPT-4.1
OpenAI
Overall
69.5
19
GLM-OCR
Zhipu AI
Overall
64.2
20
Qwen3.5-2B
Alibaba
Overall
62.6
21
Qwen3.5-0.8B
Alibaba
Overall
57.8
22
GPT-5-Nano
OpenAI
Overall
52.0
23
Llama-3.2-Vision-11B
Meta
Overall
50.8
24
Pixtral-12B
Mistral AI
Overall
46.5
0
Gemma-3-12B-IT
Overall
0.0
0
Datalab Marker
Datalab
Overall
0.0
0
Qwen-VL-OCR
Alibaba
Overall
0.0