Spatial Understanding on SpatialScore OpenSource Subset benchmark leaderboard · SOTA2 Research

Spatial Understanding on SpatialScore OpenSource Subset

87.66Overall ScoreHuman-level

Evaluation Results

Method	Links
Human-level2025.05		87.66	96.87	89.72	81.49	92.15	84.05	90.34	75.58	92.99	89.41	84.19
GPT-52025.05		59.09	78.08	57.59	56.33	54.04	56.85	70.97	42.22	67.07	47.29	62.13
Gemini-2.5-Pro2025.05		56.7	73.38	64.06	51.06	46.41	54.7	69.89	43.83	64.63	45.57	56.25
Qwen3-VL-235B-A22B2025.05		54.82	57.27	65.19	52.84	52.47	56.85	66.9	44.28	67.38	38.42	49.63
Qwen3-VL-32B2025.05		51.86	43.4	61.16	51.18	50.22	57.81	67.13	46.47	61.28	34.73	47.79
Qwen3-VL-8B2025.05		50.01	44.3	53.49	51.55	45.74	51.39	58.62	37.92	57.62	51.97	51.84
Qwen3-VL-8B2025.05		49.58	50.11	54.48	50.99	43.5	54.84	62.3	38.43	58.23	40.64	43.38
Qwen3-VL-8B2025.05		48.72	57.27	52.67	58.59	36.77	49.84	55.4	40.51	54.88	37.93	44.85
Qwen3-VL-30B-A3B2025.05		48.6	46.31	58.86	45.06	47.98	53.87	61.38	33.65	56.1	40.39	45.59
InternVL3-78B2025.05		48.23	50.34	59.19	44.46	45.74	44.78	65.98	35.74	57.32	37.93	43.75
Qwen3-VL-4B2025.05		46.74	65.55	52.24	53.87	33.86	34.52	52.18	37.27	55.49	40.89	44.85
Qwen3-VL-4B2025.05		46.49	46.53	53.46	47.36	39.01	45.15	54.48	31.86	53.05	54.93	42.28
Qwen2.5-VL-72B2025.05		45.44	53.69	49.49	55.88	36.1	47.21	60.92	32.13	53.96	37.93	22.43
Qwen3-VL-4B2025.05		45.28	56.15	54.98	42.4	36.55	50.05	58.85	28.92	57.32	31.53	38.24
Claude-4.5-Sonnet2025.05		44.64	51.01	49.67	48.32	39.91	50.02	50.57	36.31	53.35	27.59	41.18
Qwen3-VL-8B2025.05		42.97	38.26	50.35	43.58	37.67	48.61	55.86	31.45	51.52	34.48	41.54
Qwen3-VL-4B2025.05		40.39	37.81	48.22	36.81	34.75	43.7	54.71	26.92	50.61	35.96	38.24
Qwen3-VL-8B2025.05		27.81	21.7	17.21	27.71	30.49	39.38	41.61	26.33	19.21	26.11	20.59
Chance-level (Random)2025.05		26.12	23.71	22.8	20.19	31.84	24.76	34.94	21.83	25.3	26.6	28.68
Qwen3-VL-4B2025.05		24.96	27.29	11.2	25.63	25.34	37.29	27.82	22.74	23.17	22.66	20.22