Question Answering on MMLU-Pro standard (test) benchmark leaderboard · SOTA2 Research

Question Answering on MMLU-Pro standard (test)

80.8AccuracyGPT-4.1

Evaluation Results

Method	Links
GPT-4.12025.11		80.8	—	—	—
GPT-4.12025.11		80.6	71.5	0.172	0.172
GPT-4.12025.11		80.2	81.2	0.124	0.045
Qwen3-30B-A3B-Instruct2025.11		79.7	—	—	—
Qwen3-30B-A3B-Instruct2025.11		79.7	73.4	0.191	0.19
Qwen3-30B-A3B-Instruct2025.11		79.7	68.4	0.191	0.189
GPT-4.12025.11		79.6	79.8	0.141	0.092
Qwen3-30B-A3B-Instruct2025.11		79.4	72.4	0.179	0.162
Qwen3-30B-A3B-Instruct2025.11		79	75.5	0.172	0.152
DeepSeek-V32025.11		78.8	76.9	0.162	0.12
DeepSeek-V32025.11		78.8	81.7	0.135	0.059
Qwen3-30B-A3B-Instruct2025.11		78.6	77.7	0.14	0.059
DeepSeek-V32025.11		77.1	77.6	0.151	0.074
DeepSeek-V32025.11		77	—	—	—
Qwen3-4B-Instruct2025.11		72	—	—	—
Qwen3-4B-Instruct2025.11		72	74.4	0.261	0.259
Qwen3-4B-Instruct2025.11		72	53.5	0.334	0.323
Qwen3-4B-Instruct2025.11		71.6	77.2	0.236	0.23
Qwen3-4B-Instruct2025.11		71.1	76.4	0.25	0.243
Baichuan-M2-32B-GPTQ-INT42025.11		70.9	—	—	—
Baichuan-M2-32B-GPTQ-INT42025.11		70.9	56	0.287	0.288
Baichuan-M2-32B-GPTQ-INT42025.11		70.9	52	0.248	0.151
Qwen3-4B-Instruct2025.11		70.7	77.3	0.174	0.091
Baichuan-M2-32B-GPTQ-INT42025.11		70.1	73.9	0.203	0.147
Baichuan-M2-32B-GPTQ-INT42025.11		70	72	0.21	0.156
Baichuan-M2-32B-GPTQ-INT42025.11		69.8	73.4	0.185	0.07
Mistral-3.2-24B-Instruct2025.11		66.6	—	—	—
Mistral-3.2-24B-Instruct2025.11		66.6	65.8	0.299	0.294
Mistral-3.2-24B-Instruct2025.11		66.6	73.5	0.196	0.092
Mistral-3.2-24B-Instruct2025.11		65.9	72.3	0.209	0.129
Mistral-3.2-24B-Instruct2025.11		65.7	72.1	0.229	0.18
Mistral-3.2-24B-Instruct2025.11		64.8	70	0.263	0.242