Audio Understanding on MMAR (comprehensive evaluation) benchmark leaderboard · SOTA2 Research

Audio Understanding on MMAR (comprehensive evaluation)

62.4Sound ScoreMPAR2

Evaluation Results

Method	Links
MPAR22026.02		62.4	44.3	64.26	—	—	—	—	60.32
Gemini 2.0 Flash2026.01		61.21	50.97	72.11	81.82	72.48	65.85	70.83	65.6
Qwen3-Omni2026.01		59.39	54.37	70.41	90.91	74.77	63.41	70.83	65.9
Qwen2.5-Omni2026.04		58.8	40.8	59.9	—	—	—	56.7	—
UAS-Audio2026.04		58.8	45.2	66	—	—	—	60.1	—
Qwen2.5-Omni2026.01		55.76	41.75	54.42	45.45	55.96	57.32	54.17	51.4
Step-Audio22026.04		54.6	42.2	61.2	—	—	—	56.8	—
DIFFA-22026.01		54.55	41.75	53.4	45.45	58.26	54.88	37.5	50.8
GPT-4o-Audio2026.01		53.94	50.97	70.41	63.64	72.48	62.2	75	63.5
DIFFA-22026.01		53.94	36.41	52.72	36.36	56.88	53.66	45.83	50.2
Qwen2.5-Omni2026.02		53.94	46.12	53.74	—	—	—	—	53.8
Qwen2.5-Omni2026.02		53.94	42.23	58.84	—	—	—	—	55.2
MPAR22026.02		53.68	41.56	58.45	—	—	—	—	55.62
MiniCPM-O2026.01		49.7	36.41	50	36.36	53.67	45.12	54.17	48.6
Kimi-Audio2026.04		49.7	33	58.5	—	—	—	48	—
Baichuan-Omni-1.52026.01		41.21	33.01	40.48	36.36	48.62	39.02	41.67	40.7
UAS-Audio-D2026.04		40	36.89	44.56	—	—	—	43.3	—
DIFFA2026.01		37.58	31.07	39.46	36.36	43.12	45.12	25	37.2
Qwen2-Audio2026.01		33.33	24.27	32.31	9.09	31.19	30.49	25	30
Salmonn2026.01		30.91	25.73	34.35	9.09	37.61	28.05	37.5	32.8
Salmonn2026.01		30.3	29.61	34.69	9.09	34.86	35.37	41.67	33.2
GLM-4-Voice2026.04		29.7	19.9	34.35	—	—	—	29.6	—
Qwen-Audio-Chat2026.01		27.88	20.39	22.11	9.09	25.23	25.61	20.83	23.5
OpenOmni2026.01		20.61	22.33	35.37	18.18	27.06	23.17	25	27
LTU2026.01		19.39	19.9	13.95	18.18	24.77	21.95	16.67	19.2