Search-oriented Visual Question Answering on Search-oriented Benchmarks 1.0 (test val)

66.45InfoSeek ScoreGemini-3-Pro

Evaluation results

Method
Gemini-3-Pro2025.12	66.45	69.06	74.27	4,852	72.61	75.91	59.69	86
GPT-5.22025.12	65.55	67.64	66.08	4,820	68.78	78.18	65.99	80.67
SenseNova-MARS-32B2025.12	65.25	69.74	74.27	5,443	72.61	74.14	60.83	86.67
SenseNova-MARS-8B2025.12	61.7	64.2	67.84	4,164	67.11	70.19	56.22	84.67
Gemini-3-Flash2025.12	61.1	61.26	62.57	4,164	64.89	67.92	48.06	82.67
GPT-4o2025.12	59.55	55.09	49.12	3,016	66.34	63.67	40.09	76.67
Gemini-3-Pro2025.12	56.3	55.87	62.57	2,689	59.22	64.07	40.06	82
GPT-52025.12	55.95	60.12	52.63	3,836	62.61	70.58	56.02	84.67
MMSearch-R12025.12	55.1	52.49	53.8	2,033	58.4	57.4	48.4	74
Gemini-3-Flash2025.12	54.85	53.68	57.31	2,197	56.5	63.57	38.9	82.67
GPT-52025.12	54.15	50.24	35.09	2,262	54.39	61.7	44.39	79.33
Gemini-2.5-Flash2025.12	53.7	58.05	59.06	4,000	61.72	68.81	47.75	75.33
GPT-4o2025.12	52.9	42.38	23.39	1,311	48	51.73	28.18	79.33
DeepEyesV22025.12	51.1	—	63.7	—	60.6	59.4	—	—
GPT-5.22025.12	50.4	50.92	43.27	2,492	50.94	59.92	47	80
Qwen2.5-VL-32B-Instruct2025.12	50.1	53.45	49.71	3,344	52.22	65.15	42.17	81.33
Qwen3-VL-32B-Instruct2025.12	49.85	53.82	49.12	3,443	54.28	64.17	42.87	82
DeepMMSearch-R12025.12	47.51	—	—	—	—	55.87	—	—
Qwen3-VL-8B-Instruct2025.12	46.15	51.52	47.37	2,787	53.61	62.29	39.37	84
Gemini-2.5-Flash2025.12	44.1	40.87	21.64	754	43.78	55.48	31.57	82
GPT-4o-mini2025.12	42.35	45.65	38.6	2,623	50	50.84	31.54	80
Visual-ARFT2025.12	37.95	40.13	34.5	2,492	41.72	42.45	25.4	74
Qwen2.5-VL-32B-Instruct2025.12	36.65	32.01	11.7	393	30.5	48.57	21.4	71.33
GPT-4o-mini2025.12	35.95	33.08	15.79	131	36.83	44.42	24.63	72.66
Qwen2.5-VL-7B-Instruct2025.12	31.95	27.7	7.6	58	26.28	47.88	19.63	60
Qwen3-VL-32B-Instruct2025.12	28.95	35.22	16.96	1,902	32.17	45.9	31.59	72.67
Qwen2.5-VL-7B-Instruct2025.12	28.8	35.5	32.16	1,934	36	42.35	22.52	67.33
Qwen3-VL-8B-Instruct2025.12	23.15	29.24	11.7	1,213	24.22	42.94	23.18	67.33