Document Parsing on DocPTBench Chinese benchmark leaderboard · SOTA2 Research

Document Parsing on DocPTBench Chinese

30.4Overall Edit DistanceGemini2.5-Pro

Evaluation Results

Method	Links
Gemini2.5-Pro2026.02		30.4	22	52.2	82.9	18.1
dots.ocr2026.02		37.3	35.8	54.4	67.6	31.8
Dolphin-v22026.02		37.3	35.8	54.4	53.6	29.2
Kimi-VL2026.02		38.7	22	52.2	67.8	18.1
PaddleOCR-VL2026.02		39.6	37.7	52.6	65.3	37.9
Qwen-VL-Max2026.02		42.7	41.5	57.2	71.6	34.4
OCRFlux2026.02		45.8	40.4	81.1	54.3	32.1
olmOCR2026.02		46.1	27.2	66.9	56.9	24.4
Nanonets-OCR22026.02		46.1	44.6	76.4	66	19.5
MinerU2.52026.02		47.4	53.6	62	59.8	40.3
GLM-4.5v2026.02		49.6	47.7	66.2	54	35.7
Nanonets-OCR2026.02		52.1	42	67	50.6	32.7
MonkeyOCR2026.02		52.8	43.9	61.6	37.4	44.1
Doubao-1.6-v2026.02		55.4	58.2	61.1	37.9	40.2
Qwen2.5-VL-72B2026.02		57	56.6	61.8	55.5	51.3
DeepSeek-OCR2026.02		57.8	57.6	74.1	35.4	40.4
Dolphin2026.02		71.5	71.5	82.8	19.3	57.7
SmolDocling2026.02		93.7	99.2	99.9	2.4	75.9