List your company
ResearchFormerly WizWand

Mathematical Reasoning on AIME 2026 (avg@32)

59.1Average Success Rate (avg@32)CTPO
1.17216.21131.2546.289May 8, 2026May 9, 2026May 10, 2026May 11, 2026May 12, 2026May 13, 2026
Updated 5mo ago

Evaluation results

MethodLinks
CTPOBase Model=Qwen3-14B, Reasoning Setting=Tool-integrated2026.0559.1
GSPOBase Model=Qwen3-14B, Reasoning Setting=Tool-integrated2026.0556.1
CTPOBase Model=Qwen3-4B, Reasoning Setting=Tool-integrated2026.0552.1
GRPOBase Model=Qwen3-14B, Reasoning Setting=Tool-integrated2026.0549.8
GSPOBase Model=Qwen3-4B, Reasoning Setting=Tool-integrated2026.0547.6
ConSPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0546.8
DisCOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0545.1
GRPOBase Model=Qwen3-4B, Reasoning Setting=Tool-integrated2026.0544.1
GRPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0543.9
Dr. GRPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0543.3
DAPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0537.9
Base ModelBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0535.5
ConSPOBackbone=DeepSeek-R1-Distill-Llama-8B2026.0535.5
GRPOBackbone=DeepSeek-R1-Distill-Llama-8B2026.0534.2
DisCOBackbone=DeepSeek-R1-Distill-Llama-8B2026.0532.8
Dr. GRPOBackbone=DeepSeek-R1-Distill-Llama-8B2026.0532.4
DAPOBackbone=DeepSeek-R1-Distill-Llama-8B2026.0529.9
ConSPOBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0523.9
GMPOBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0523.8
SAPOBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0523.3
CISPOBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0523
DisCOBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0521.7
DAPOBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0520.8
Base ModelBackbone=DeepSeek-R1-Distill-Llama-8B2026.0520.6
GRPOBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0520.4
Dr. GRPOBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0519.5
Base ModelBase Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0513.9
BaseBase Model=Qwen3-14B, Reasoning Setting=Tool-integrated2026.056.9
BaseBase Model=Qwen3-4B, Reasoning Setting=Tool-integrated2026.053.4