Comparing parsers on some tags
Pass rate minus median of the reported systems.
table
math
form
multi column
multilingual
handwriting
rotated
tiny
degraded
list
heading
weights
Datalab accurate
closed
+11.3
+4.5
+13.7
+13.4
+25.7
+21.8
+29.8
+15.2
+22.1
+17.2
+17.3
Gemini 3.8 Flash*
closed
+11.5
+5.4
+12.3
+11.3
+24.5
+22.4
+28.3
+15.2
+21.4
+14.2
+17.0
Reducto
closed
+8.3
+4.1
+10.0
+8.7
+5.1
+10.6
+19.2
+14.0
+16.3
+5.7
+12.1
LlamaParse
closed
+9.6
+0.4
+4.9
+10.0
+20.8
+16.3
+26.9
+10.6
+15.3
+13.1
+12.8
chandra-ocr-2
open
+5.0
+3.6
+5.3
+8.7
+15.1
+11.3
+8.3
+14.4
+18.2
+14.1
+14.8
Mistral OCR
closed
+8.3
-0.2
+5.4
+4.7
+15.6
+8.5
+24.3
-7.0
+5.4
-2.8
+8.1
Claude Sonnet 5.5*
closed
+11.2
+6.6
+14.7
+8.6
+22.6
+18.9
+28.1
+16.4
+21.0
+15.6
+11.8
dots.mocr
open
+5.3
+1.3
-3.9
+2.8
+17.6
-1.8
+13.8
+1.3
-2.4
+5.0
+7.5
GPT-5.6 Sol
closed
+8.3
-0.3
+7.2
+6.1
+14.1
+11.2
+12.7
+10.9
+14.7
+9.5
+13.8
MinerU2.5-Pro
open
+8.6
+1.8
+3.1
+1.9
-7.9
-17.4
+19.7
+8.7
-2.1
+0.1
-5.1
Infinity-Parser2-Pro
open
+1.9
+3.4
+3.6
+7.1
+1.1
+15.9
-0.1
+8.7
+16.7
+3.7
+3.0
Extend
closed
+9.9
-51.6
+11.0
+2.9
+9.0
+17.0
+23.7
+11.5
+14.6
+2.9
+5.0
Surya OCR 2
open
-6.3
+1.8
-5.1
-1.1
+15.0
-5.8
-17.2
-1.3
-6.0
+6.2
+9.7
Azure
closed
-0.4
-4.4
+3.4
-0.6
-5.8
-16.7
+15.5
+9.8
-7.3
-0.1
-3.8
LightOnOCR-3-4B
open
-8.1
+4.5
+6.9
+2.8
-1.1
+11.6
-1.3
+9.3
+15.6
+3.0
+7.0
LightOnOCR-3-0.8B
open
-7.3
+4.1
+5.9
+0.2
-7.1
+1.8
+0.1
+6.2
+7.1
+0.1
+0.2
GLM-OCR
open
+1.3
+0.2
-10.0
-1.8
-21.9
-10.8
-0.9
-10.2
-3.1
-14.2
-16.4
Nemotron 3.5 Parse
open
-4.4
-13.0
-3.7
-2.7
+4.0
-13.8
-15.0
-6.4
-8.2
-8.2
-0.2
Unlimited-OCR
open
-6.3
-0.7
-8.8
-7.4
-12.5
-21.7
-44.1
-16.4
-8.0
-15.1
-9.8
PaddleOCR-VL-1.6
open
+0.4
-25.8
-10.6
-9.2
-10.6
-13.2
+7.6
-17.0
-13.1
-18.6
-10.7
Infinity-Parser2-Flash
open
-13.8
+3.7
-3.1
-0.2
-11.5
+7.7
-27.0
-3.3
+10.8
-3.9
-5.4
DeepSeek-OCR-2
open
-12.1
-7.9
-18.8
-10.7
-8.4
-24.2
-49.1
-23.9
-19.0
-15.3
-5.1
Indra OCR
open
-20.3
-32.6
-20.4
-15.0
+8.4
-43.4
-12.4
-4.7
-31.5
-3.1
-3.1
HunyuanOCR-1.5
open
-18.9
-51.1
-21.0
-14.6
-10.4
-15.2
-19.0
-24.2
-10.0
-21.1
-8.7
olmOCR-2
open
-24.6
-5.3
-8.4
-11.9
-23.5
+2.6
-6.9
-20.9
+2.1
-11.6
-11.1
Nanonets-OCR2-3B
open
-19.3
-13.1
-4.5
-16.8
-18.6
-4.1
-35.4
-32.2
-3.8
-26.8
-6.8
median
83.3
89.6
78.6
80.3
61.2
59.0
64.6
78.4
63.3
78.7
69.6
A test counts toward every tag it has; multilingual is any language but English. The last row is the median itself.
* Used in creating and verifying at least some tests.