Text Spotting on Diverse image domains benchmark leaderboard · SOTA2 Research

Text Spotting on Diverse image domains

71.4Overall PerformanceHunyuanOCR-1.5

Evaluation Results

Method	Links
HunyuanOCR-1.52026.07		71.4	53.21	79.43	75.84	78.4	75.03	65.22	74.51	65.12	76.09
HunyuanOCR2026.07		70.92	56.76	73.63	73.54	77.1	75.34	63.51	76.58	64.56	77.31
PaddleOCR-VL-1.62026.07		61.95	41.36	72.2	58.56	70.61	65.24	61.85	63.63	54.6	69.52
BaiduOCR2026.07		61.9	38.5	78.95	59.24	59.06	66.7	63.66	68.18	55.53	67.38
Qwen3.5 A17B2026.07		59.76	44.92	52.56	58.16	71.54	67.42	55.98	62.58	56.11	68.56
Gemini 3.1 Pro2026.07		59.53	46.83	54.89	62.62	63.37	63.96	54.53	64.29	55.3	70.02
Seed-2.0-Vision2026.07		56.32	44.77	45.85	61.7	66.89	61.87	55.73	52.05	46.53	71.49
Qwen3-VL-235B-A22B-Ins.2026.07		53.62	46.15	43.78	48	68.9	64.01	47.53	45.91	54.56	63.79
PaddleOCR2026.07		53.38	32.83	70.23	51.59	56.39	57.38	50.59	63.38	44.68	53.35
Qwen3-VL-2B-Ins.2026.07		29.68	29.43	19.37	20.85	50.57	35.14	24.42	12.13	34.9	40.1
Gemini 2.5 Pro2026.07		23.44	21.79	35.16	10.02	38.49	29.89	20.8	17.59	18.33	18.9