Mathematical and Spatial Reasoning on MathSpatial-Bench benchmark leaderboard · SOTA2 Research

Mathematical and Spatial Reasoning on MathSpatial-Bench

100C3MHuman

Evaluation Results

Method	Links
Human2026.02		100	89.4	97.6	98.2	96.1	97	95.2	96.5	94.4	97.7	96.2	96.3
Gemini-2.5-flash2026.02		34.3	42.5	46.7	26.7	58.5	47.8	43.8	50	25	57.4	33.3	48.5
Gemini-2.5-pro2026.02		31.4	42.5	43.5	26.7	58.5	44.9	28.1	45.6	41.7	49.6	33.3	44.9
GPT-52026.02		28.6	53.7	66.2	48.2	61.6	44.8	71.7	52.3	60	68.1	57.7	58.5
Llama-4-Maverick-17B-128E2026.02		24.5	32.5	41	25	29.1	4.5	47.6	4.6	15.2	29.6	30.8	23.8
MathSpatial-Llama3-8B2026.02		16.3	30.1	24.5	41.1	20.4	30.6	27.6	10.2	20.8	16.4	7.7	20.3
Llama3-8B2026.02		10.2	15.4	20.7	17.9	15.1	20.9	23.4	9.3	13.6	12.2	7.7	15
Claude 42026.02		6.1	33.3	46.2	21.4	30.8	1.5	60.7	0.2	32	41.3	53.8	26.7
Pixtral-12B2026.02		4.1	22.8	34.1	26.8	20.4	1.5	34.5	0	8	12.2	3.8	15.3
GLM-4.5V2026.02		4.1	32.5	43.4	30.4	23	1.5	57.9	0	22.4	14.1	30.8	21
MathSpatial-Qwen2.5-VL-7B2026.02		4.1	40.7	46.2	30.4	27.7	0.7	46.2	0	21.6	17.4	26.9	22.1
GPT-4.12026.02		2	45.5	43.1	26.8	24.1	0	55.2	0	29.6	18.3	46.2	22.6
GPT-4o2026.02		2	36.6	34.1	33.9	23.5	1.5	45.5	0	26.4	16.4	26.9	19.6
Claude 3.72026.02		2	36.6	39.3	21.4	21.6	1.5	57.9	0	23.2	26.3	34.6	21.5
Claude 3.52026.02		2	32.5	42.8	23.2	26.9	0.7	57.2	0	24.8	21.1	46.2	22.3
Kimi-VL-A3B-Thinking2026.02		2	33.3	31.7	50	23.8	0	47.6	0.6	23.2	5.6	11.5	18.2
Qwen2.5-VL-72B-Instruct2026.02		2	33.3	35.9	23.2	26.9	0	47.6	0	20.8	17.8	19.2	19.7
MathSpatial-InternVL3-8B2026.02		2	50.4	36.2	33.9	22.7	11.9	56.6	3.9	23.2	15.5	15.4	22.6
Qwen2.5-VL-7B2026.02		0	32.5	34.8	17.9	24.4	0	41.4	0	20	13.6	15.4	17.8
DeepSeek-VL2-7B2026.02		0	14.6	11.7	7.1	5.9	4.5	2.1	2.1	11.2	4.2	7.7	6.6
InternVL3-8B2026.02		0	29.3	35.9	23.2	21.3	0.7	40	0	19.2	14.6	15.4	17.4
Idefics3-8B2026.02		0	22.8	25.9	14.3	18.2	1.5	22.1	0	8	6.1	0	11.7