Visual Reasoning on VisualFLIP benchmark leaderboard · SOTA2 Research

Visual Reasoning on VisualFLIP

85.6Cardinality AccuracyGPT-5.5

Evaluation Results

Method	Links
GPT-5.52026.06		85.6	3.7	86.4	5.7	68	7.8	65.3	6.6	78.6	5.8
Gemini 3.5 Flash2026.06		84.9	7.1	90.5	4.5	70.7	11.4	68.6	9.5	81.2	7.3
Qwen3.6-Plus2026.06		83.6	8.3	90.5	3.4	69.3	12.8	66.1	5.3	80.2	6.8
Gemini 3.1 Pro2026.06		79.5	8.6	88.3	5.9	68.7	16	59.3	15.5	77.1	10.1
Qwen3.5-Flash2026.06		74	6.4	87.2	3.8	52.7	17.3	59.3	7.9	72.1	7.8
Seed 2.0 Mini2026.06		61.7	10.1	86.2	8.8	67.4	18.4	40.2	14.5	68.6	11.8
Kimi K2.6-1T2026.06		48.6	2.9	39.9	1.7	41.6	2.4	35.6	5.6	41.6	2.9
MiMo-v2.5-310B2026.06		47.3	8.8	59	7.6	46.2	15.7	43.2	8.6	51	9.6
Claude Opus 4.72026.06		43.2	16	74	8.3	48.7	20.3	46.6	12.2	57.2	13.1
GLM-5V-Turbo2026.06		41.1	13.3	78.7	5.5	53.6	17.3	46.6	11.1	59.7	10.2
GLM-4.6V-106B2026.06		37	14	42.9	32.5	44.7	32.3	23.7	14.3	38.7	26.6
Grok 4.32026.06		33.6	22.9	73.6	12	40	28.8	43.2	17.2	52.5	18.4
GPT-5-mini2026.06		28.1	27.3	53.8	29	47.3	30.4	44.1	20.2	45.3	27.6
Qwen3-VL-32B2026.06		27.4	50.5	16.8	67.4	42	36.6	19.5	37.5	25	51.4
Qwen3-VL-235B2026.06		26.7	47.2	19	60.1	38.7	39.7	28	24.2	26.5	47.7
Claude Opus 4.62026.06		19.2	25	38.1	17.3	30.7	29.1	14.4	32.7	28.4	23.2
Qwen3-VL-8B2026.06		15.8	55.1	11.7	61.5	28	43.2	20.3	48.5	17.6	52.7
GPT-4o2026.06		13	57.1	20.9	61.7	40	34.7	20.3	34.9	23.3	50.6
CoF-7B2026.06		7.5	33.3	13.2	54.7	14.7	42	3.4	49	10.6	47.4
DeepEyesV2-7B2026.06		7.5	34.4	8.4	50.7	11.3	41.5	10.2	32.8	9.2	42.6
PixelReasoner-8B2026.06		4.8	38.1	12.1	55.1	5.3	56.2	5.1	50	7.9	51.6
Mini-o3-7B2026.06		4.8	38.3	10.6	49.6	5.3	42.1	5.1	42	7.3	44.7
DeepEyes-7B2026.06		4.1	42.1	8.1	64.6	8	45.6	7.6	44.3	7.1	53
Qwen2.5-VL-7B2026.06		3.4	45.2	12.8	61.2	6.7	53.6	11	35.2	9.2	53