SOTA Science Reasoning on GPQA-D (Acc, ∆Tok) and PapersWithCode

76.3Accuracy

Qwen3-Next-80B

Updated 8d ago

Evaluation Results

Method
Qwen3-Next-80B 2025.10	76.3	-
Qwen3-Next-80B + THINKBRAKE 2025.10	72.7	-25.6
Qwen3-4B-Thinking + NoWait 2025.10	68.2	-19.6
Qwen3-4B-Thinking + DEER 2025.10	68.2	-1.3
Phi-4-Reasoning + THINKBRAKE 2025.10	67.7	-1.1
Qwen3-32B + THINKBRAKE 2025.10	65.7	-19.9
Qwen3-32B 2025.10	65.2	-
Qwen3-4B-Thinking 2025.10	64.1	-
Phi-4-Reasoning + NoWait 2025.10	63.1	-19.1
Qwen3-4B-Thinking + Dynasor-CoT 2025.10	62.6	-22.6
Qwen3-4B-Thinking + THINKBRAKE 2025.10	62.6	-32.7
Phi-4-Reasoning 2025.10	62.6	-
Qwen3-14B + NoWait 2025.10	62.1	-20.5
Qwen3-14B + THINKBRAKE 2025.10	61.1	-39.9
Qwen3-14B 2025.10	60.6	-
Qwen3-32B + NoWait 2025.10	60.1	-12.4
Phi-4-Reasoning + THINKBRAKE-p 2025.10	59.1	-62.8
Qwen3-4B + NoWait 2025.10	57.6	-26.6
Qwen3-14B + THINKBRAKE-p 2025.10	57.6	-66.4
Qwen3-4B + THINKBRAKE 2025.10	55.6	-29.7
Qwen3-32B + ThinkLess 2025.10	54.5	-100
Phi-4-Reasoning + ThinkLess 2025.10	54	-100
Qwen3-32B + THINKBRAKE-p 2025.10	53.5	-51
MCPO-DAPO 2026.05	52.44	-
Qwen3-4B 2025.10	51	-
MCPO-GRPO 2026.05	50.57	-
MGS 2026.05	50.25	-
DAPO 2026.05	50.11	-
Qwen3-4B-Thinking + ThinkLess 2025.10	50	-100
MT-GRPO 2026.05	49.89	-
CLIPO 2026.05	49.83	-
Qwen3-14B + ThinkLess 2025.10	49.5	-100
MCPO-DAPO 2026.05	49.13	-
MCPO-GRPO 2026.05	48.81	-
GRPO 2026.05	48.44	-
DeepSeek-R1-7B 2025.10	48	-
Qwen3-4B + THINKBRAKE-p 2025.10	47.5	-54.2
Qwen3-4B-Thinking + THINKBRAKE-p 2025.10	47	-67.5
DAPO 2026.05	46.37	-
Qwen3-4B + ThinkLess 2025.10	46	-100
DeepSeek-R1-7B + THINKBRAKE 2025.10	45.5	-51.5
MGS 2026.05	45.4	-
MT-GRPO 2026.05	44.06	-
GRPO 2026.05	43.95	-
DeepSeek-R1-7B + DEER 2025.10	43.9	-16.7
DeepSeek-R1-7B + NoWait 2025.10	43.4	-35.5
CLIPO 2026.05	42.68	-
Base Model 2026.05	42.07	-
DeepSeek-R1-7B + THINKBRAKE-p 2025.10	36.4	-76.8
Base Model 2026.05	35.45	-
DeepSeek-R1-7B + Dynasor-CoT 2025.10	32.3	-71.2
DeepSeek-R1-7B + ThinkLess 2025.10	30.3	-100