Models
23 models evaluated across 3 document AI benchmarks.
1
Gemini 3.1 Pro
Overall
83.2
2
Nanonets OCR2+
Nanonets
Overall
81.8
3
Gemini-3-Pro
Overall
81.4
4
GPT-5.4
OpenAI
Overall
81.0
5
Claude Sonnet 4.6
Anthropic
Overall
80.8
6
Claude Opus 4.6
Anthropic
Overall
80.3
7
Gemini-3-Flash
Overall
79.9
8
GPT-5.2
OpenAI
Overall
79.2
9
Qwen3.5-9B
Alibaba
Overall
77.0
10
Qwen3.5-4B
Alibaba
Overall
73.1
11
Mistral Small 4
Mistral AI
Overall
71.5
12
GPT-5-Mini
OpenAI
Overall
70.8
13
GPT-4.1
OpenAI
Overall
70.0
14
Claude Haiku 4.5
Anthropic
Overall
69.6
15
Ministral-8B
Mistral AI
Overall
69.3
16
GLM-OCR
Zhipu AI
Overall
63.6
17
Qwen3.5-2B
Alibaba
Overall
63.2
18
Qwen3.5-0.8B
Alibaba
Overall
58.0
19
GPT-5-Nano
OpenAI
Overall
50.7
20
Llama-3.2-Vision-11B
Meta
Overall
50.1
21
Pixtral-12B
Mistral AI
Overall
46.0
0
Gemma-3-12B-IT
Overall
0.0
0
Datalab Marker
Datalab
Overall
0.0