Multilingual Reasoning on GlobalMMLU benchmark leaderboard · SOTA2 Research

Multilingual Reasoning on GlobalMMLU

88.77English AccuracyQwen2.5-72B-Instruct

Evaluation Results

Method	Links
Qwen2.5-72B-Instruct2025.12		88.77	3.11	85.84	3.5	78.29	7.96	81.48	4.51	82.78	3.75	81.8	9.63	87.31	3.3	81.05	6.16	83.67	4.74	82.39	6.09
Llama-3.3-70B-Instruct2025.12		88.62	2.81	85.11	4.24	79.89	6.92	80.25	5.94	83.92	4.91	78.87	11.72	86.87	3.52	80.84	7.21	83.63	4.91	81.64	7.99
Qwen3-32B-thinking2025.12		88.12	0.91	85.88	1.71	81.13	3.23	81.48	3.15	83.78	2.38	81.06	8.5	87.02	1.31	81.91	4.13	84.3	2.2	82.57	5.11
Qwen2.5-32B-Instruct2025.12		87.14	2.96	83	2.64	69.96	7.8	76.86	5.77	77.38	4.46	77.01	13.3	85.15	2.81	75.42	7.44	79.93	4.58	77.22	8.17
Llama-4-Scout-Instruct2025.12		87.06	2.61	84.33	3.99	81.31	6.87	78.71	6.12	83.95	3.71	80.33	12.57	85.7	3.3	81.11	7.25	82.93	4.86	82.2	8
Qwen3-32B2025.12		86.98	2.16	83.76	2.28	78.51	4.44	79.59	4.04	80.94	2.28	78.4	9.81	85.38	2.22	79.36	5.14	82.22	3.23	79.68	6.04