Medical Multi-task Visual Reasoning on MedAD-38K (test) benchmark leaderboard · SOTA2 Research

Medical Multi-task Visual Reasoning on MedAD-38K (test)

98.87Anatomy ID AccuracyMedAD-R1

Evaluation Results

Method	Links
MedAD-R12026.02		98.87	78.24	55.9	97.14	79.49	85.15
Lingshu2026.02		96.18	59.57	31.75	76	32.09	70.91
HuatuoGPT-Vision2026.02		95.65	55.78	33.79	95.08	59.58	75.56
GLM-4.1V-Thinking2026.02		94.87	52.05	27.6	96.59	58.78	74.02
Qwen3VL2026.02		94.68	56.67	32.8	95.29	54.6	75.44
InternVL3.52026.02		94.02	59.27	27.84	92.19	58.58	74.46
Qwen2.5VL-72B2026.02		93.75	59.24	37.46	94.32	53.43	76.25
Grok4-Fast2026.02		93.47	59.94	24.9	92.09	39.49	77
InternVL3.52026.02		93.39	58.3	31.57	94.25	71.52	75.16
Qwen2.5-VL-3B2026.02		93.2	51.27	24.57	91.78	65.3	71.41
Llama-3.2-Vision2026.02		92.38	47.29	22.22	95.21	50.33	70.59
MedVLM2026.02		91.17	48.13	30.37	93.45	74.63	71.23
MiniCPM-V-4.52026.02		91.16	56.43	16.31	92.95	54.6	71.48
ERNIE-4.5-VL2026.02		90.49	48.66	24.19	91.68	52.13	69.73
MiMo-VL-RL2026.02		88.43	57.35	32.98	90.4	50.37	72.44
LLaVA-Med2026.02		86.45	45.82	20.81	81.63	39.55	64.36
Qwen2.5-VL-7B2026.02		83.98	55.71	35.62	88.45	48.9	70.48