Multi-video understanding on MVX-Bench benchmark leaderboard · SOTA2 Research

Multi-video understanding on MVX-Bench

52.7Average ScoreSAMA

Evaluation Results

Method	Links
SAMA2026.03		52.7	41.2	53.3	78	64.6	57.3	54	78.8	32	56.7	49.3	30
GPT-5.22026.03		50.9	32	68	79.3	73.3	58	57.5	81.2	24	52.7	22.4	28.7
GPT-4o2026.03		47.7	26	49	71.3	70.7	40	70.8	76.8	23.3	52	39.3	26.7
InternVL3.5-38B2026.03		47.2	28.7	45	71.3	53.3	58	53.8	82.6	22.7	46	50.4	28.7
Qwen2.5-VL-32B2026.03		46.3	20.7	52	72	50.7	56.7	58.5	78.3	19.3	51.3	45.3	27.3
Gemma-3-27B-IT2026.03		44.7	28	46	72.7	56	56.7	33	82.6	16.7	52.7	35.9	27.3
InternVL3.5-8B2026.03		44.7	30.7	49	68.7	49.3	60	26.4	81.2	21.3	51.3	43.6	26
InternVL3.5-4B2026.03		44.3	26.7	55	64.7	34.7	57.3	48.1	79.7	25.3	53.3	44.4	22
InternVL3.5-14B2026.03		43.8	26.7	51	68	44	62	33	79.7	22.7	46	40.2	26
Gemma-3-12B-IT2026.03		41.5	25.3	43	65.3	49.3	60	16	71	23.3	52	36.3	22
Gemma-3-4B-IT2026.03		40.6	28.7	49	59.3	30.7	54	16	72.5	26.7	56	42.7	24
Qwen2.5-VL-3B2026.03		38.1	29.3	55	45.3	27.3	56	34.9	63.8	16	47.3	41	22.7
Pixtral-12B2026.03		35.4	23.3	43	27.3	33.3	58	27.4	36.2	30.7	46.7	39.3	25.3
Phi-4-Multimodal2026.03		34.4	21.3	55	28.7	23.3	55.3	34.9	21.7	27.3	48	42.6	22.7
Idefics3-8B-Llama32026.03		33.4	25.3	51	31.3	24.7	54.7	25.5	27.5	18.7	50.7	35	23.3
Random2026.03		33.2	25	50	25	25	50	25	25	25	50	40.5	25
Bee-8B2026.03		33.2	26	55	28.7	23.3	52.7	34.9	24.6	26.7	32.7	45.1	22
LLaVA-OneVision-7B2026.03		32.7	26	47	29.3	26	54	22.6	23.2	18	52	40.5	19.3
Qwen2.5-VL-7B2026.03		32.6	22	59	24.7	21.3	58.7	21.7	23.2	22	44.7	44.4	20
MiniCPM-V-22026.03		32	28.7	41	24	23.3	51.3	22.6	15.9	28	49.3	38.5	22
Phi-3.5-Vision2026.03		22.8	22.2	13	28.1	8.8	58	19.8	12.5	18.8	0	9.1	25