List your company

Mathematical Reasoning on OlympiadBench (Accuracy, Think ↓, Route%)

71.5AccuracyAT
48.93254.79160.6566.509Jun 22, 2026

Evaluation results

MethodLinks
ATBackbone=Qwen3-8B2026.0671.5——
SC-RouteBackbone=Qwen3-8B2026.0670.4——
DARTBackbone=Qwen3-8B2026.0669.84552.1
DARTBackbone=DeepSeek-V3.2, Hosted API=true2026.0669.132—
SC-RouteBackbone=DeepSeek-V3.2, Hosted API=true2026.0667.1——
ATBackbone=DeepSeek-V3.2, Hosted API=true2026.0666.1——
NTBackbone=DeepSeek-V3.2, Hosted API=true2026.0663.6——
SC-RouteBackbone=Qwen3-14B2026.0662——
DARTBackbone=Qwen3-14B2026.06621558
SC-RouteBackbone=Qwen3-32B2026.0658.5——
DARTBackbone=Qwen3-32B2026.0658.51659.5
ATBackbone=Qwen3-32B2026.0654——
ATBackbone=Qwen3-14B2026.0653——
NTBackbone=Qwen3-14B2026.0651.5——
NTBackbone=Qwen3-32B2026.0650.5——
NTBackbone=Qwen3-8B2026.0649.8——