List your company
ResearchFormerly WizWand

Inference Runtime on AIME 25

1Total Inference Runtime (s)DAST
0.82.153.54.85Apr 3, 2026
Updated 7mo ago

Evaluation results

MethodLinks
DASTModel=Qwen3-8B-Thinking2026.041
RL +length penaltyModel=Qwen3-8B-Thinking2026.041
S-GRPOModel=Qwen3-8B-Thinking2026.041
REDModel=Qwen3-8B-Thinking2026.041
Think or NotModel=DeepSeek-R1-Qwen-7B2026.041
DASTModel=DeepSeek-R1-Qwen-7B2026.041
RL +length penaltyModel=DeepSeek-R1-Qwen-7B2026.041
S-GRPOModel=DeepSeek-R1-Qwen-7B2026.041
REDModel=DeepSeek-R1-Qwen-7B2026.041
S-GRPOModel=DeepSeek-R1-Qwen-32B2026.041
RL +length penaltyModel=DeepSeek-R1-Llama-8B2026.041
S-GRPOModel=DeepSeek-R1-Llama-8B2026.041
REDModel=DeepSeek-R1-Llama-8B2026.041
VanillaModel=Qwen3-8B-Thinking2026.042
Think or NotModel=Qwen3-8B-Thinking2026.042
DASTModel=Qwen3-32B-Thinking2026.042
REDModel=Qwen3-32B-Thinking2026.042
VanillaModel=DeepSeek-R1-Qwen-7B2026.042
RL +length penaltyModel=DeepSeek-R1-Qwen-32B2026.042
REDModel=DeepSeek-R1-Qwen-32B2026.042
VanillaModel=DeepSeek-R1-Llama-8B2026.042
Think or NotModel=DeepSeek-R1-Llama-8B2026.042
DASTModel=DeepSeek-R1-Llama-8B2026.042
RL +length penaltyModel=Qwen3-32B-Thinking2026.043
S-GRPOModel=Qwen3-32B-Thinking2026.043
DASTModel=DeepSeek-R1-Qwen-32B2026.043
S-GRPOModel=DeepSeek-R1-Llama-70B2026.043
VanillaModel=Qwen3-32B-Thinking2026.044
VanillaModel=DeepSeek-R1-Qwen-32B2026.044
Think or NotModel=DeepSeek-R1-Qwen-32B2026.044
REDModel=DeepSeek-R1-Llama-70B2026.044
Think or NotModel=Qwen3-32B-Thinking2026.045
DASTModel=DeepSeek-R1-Llama-70B2026.045
RL +length penaltyModel=DeepSeek-R1-Llama-70B2026.045
VanillaModel=DeepSeek-R1-Llama-70B2026.046
Think or NotModel=DeepSeek-R1-Llama-70B2026.046