Complex Reasoning and Semantics on SO-Bench benchmark leaderboard · SOTA2 Research

Complex Reasoning and Semantics on SO-Bench

45.16CM ScoreSO-7B(MIX)

Evaluation Results

Method	Links
SO-7B(MIX)2026.06		45.16	29.41	28.23	33.02	77.09
SO-7B-iv2026.06		43.55	50	22.52	25.38	76.03
BAT2026.06		43.55	32.35	25.53	16.87	98.4
SO-7B-neuiv2026.06		41.94	23.53	24.32	33.6	76.98
SO-30B2026.06		41.94	32.35	27.02	32.71	71.77
SO-7B-zs2026.06		38.24	20.59	22.52	10.21	77.11
Gemini-3-pro2026.06		35.9	35.29	18.6	9.21	73.46
Gemini-2.5-pro2026.06		24.19	38.24	4.5	4.71	81.19
Qwen-2.5-Omni2026.06		22.58	29.41	10.51	11.41	85.69
Gemini-2.5-flash2026.06		22.58	38.24	3.3	8.31	88.85
Phi-4-MM2026.06		20.97	23.53	1.5	7.66	84.38
GPT-audio2026.06		17.95	23.53	0	4.64	96.1
Qwen-3-Omni2026.06		17.74	26.47	0	9.52	89.16
Kimi-Audio2026.06		17.74	26.47	9.01	7.17	74.48
Audio Flamingo 32026.06		16.13	23.53	5.41	12.31	77.85
SO-7B-so2026.06		14.52	20.59	15.62	6.61	98.73