Share your thoughts, 1 month free Claude Pro on usSee more

Math Reasoning on AIME 2024 (Acc, ∆Tok)

90Accuracy

Qwen3-Next-80B

Updated 1mo ago

Evaluation Results

Method	Links
Qwen3-Next-80B 2025.10		90	-
Qwen3-Next-80B + THINKBRAKE 2025.10		86.7	-24.3
Qwen3-14B + THINKBRAKE 2025.10		77.9	-7.4
Qwen3-4B-Thinking 2025.10		77.1	-
Qwen3-32B + THINKBRAKE 2025.10		77.1	-6.2
Qwen3-32B 2025.10		75.8	-
Phi-4-Reasoning + THINKBRAKE 2025.10		73.3	-6.4
Qwen3-14B 2025.10		71.7	-
Qwen3-14B + NoWait 2025.10		70	-19.8
Qwen3-4B 2025.10		68.8	-
Phi-4-Reasoning 2025.10		67.5	-
Qwen3-4B-Thinking + THINKBRAKE 2025.10		67.2	-33.5
Qwen3-4B-Thinking + NoWait 2025.10		66.7	-30.5
Qwen3-32B + NoWait 2025.10		66.7	-12
Qwen3-4B + THINKBRAKE 2025.10		64.6	-12.7
Qwen3-4B-Thinking + DEER 2025.10		63.3	-34.3
Qwen3-4B-Thinking + Dynasor-CoT 2025.10		63.3	-28.8
Phi-4-Reasoning + NoWait 2025.10		63.3	-9.4
Qwen3-4B + NoWait 2025.10		56.7	-16.6
DeepSeek-R1-7B + DEER 2025.10		50	-28.2
DeepSeek-R1-7B 2025.10		49.5	-
DeepSeek-R1-7B + THINKBRAKE 2025.10		45.2	-25
DeepSeek-R1-7B + NoWait 2025.10		40	-32.4
Phi-4-Reasoning + THINKBRAKE-p 2025.10		39.6	-63.1
Qwen3-32B + THINKBRAKE-p 2025.10		39.2	-61.5
Qwen3-14B + THINKBRAKE-p 2025.10		38.8	-68
Phi-4-Reasoning + ThinkLess 2025.10		36.7	-100
Qwen3-4B-Thinking + THINKBRAKE-p 2025.10		35	-72.6
DeepSeek-R1-7B + Dynasor-CoT 2025.10		33.4	-31.8
Qwen3-4B + THINKBRAKE-p 2025.10		29.2	-64.5
DeepSeek-R1-7B + THINKBRAKE-p 2025.10		25.4	-71.4
Qwen3-4B + ThinkLess 2025.10		23.3	-100
Qwen3-14B + ThinkLess 2025.10		23.3	-100
Qwen3-32B + ThinkLess 2025.10		23.3	-100
DeepSeek-R1-7B + ThinkLess 2025.10		16.7	-100
Qwen3-4B-Thinking + ThinkLess 2025.10		3.3	-100