MMLU-Redux — Results

Rank	Model	Score
1	minicpm5-2b	84.7
2	qwen3-7-max	0.95
3	qwen3-5-397b-a17b	0.949
4	qwen3-6-plus	0.945
5	qwen3-7-plus	0.945
6	kimi-k2-thinking-0905	0.944
7	qwen3-5-122b-a10b	0.94
8	qwen3-235b-a22b-thinking-2507	0.938
9	qwen3-vl-235b-a22b-thinking	0.937
10	qwen3-6-27b	0.935
11	deepseek-r1-0528	0.934
12	qwen3-5-35b-a3b	0.933
13	qwen3-6-35b-a3b	0.933
14	qwen3-5-27b	0.932
15	qwen3-235b-a22b-instruct-2507	0.931
16	kimi-k2-instruct-0905	0.927
17	kimi-k2-instruct	0.927
18	qwen3-next-80b-a3b-thinking	0.925
19	qwen3-vl-235b-a22b-instruct	0.922
20	qwen3-vl-32b-thinking	0.919
21	deepseek-v3-1	0.918
22	qwen3-5-9b	0.911
23	qwen3-next-80b-a3b-instruct	0.909
24	qwen3-vl-30b-a3b-thinking	0.909
25	qwen3-vl-32b-instruct	0.898

M

MMLU-Redux — Phases

1 phaseActive

Phase 1·

MMLU-Redux

Aug 3, 2026

Current