List your company

Mathematical Reasoning on GSM8K (test) (Acc, Cost, A/C)

95.2AccuracyGPT-4.1-mini
84.887.590.292.9May 7, 2026

Evaluation results

MethodLinks
GPT-4.1-miniType=Target Model Only2026.0595.20.029332.4
SpecReasonExt. Models=false2026.0594.90.09949.55
STEERExt. Models=false2026.0594.50.060815.54
policy-guided stepwise model routingExt. Models=false2026.0594.50.020546.09
RSDExt. Models=true2026.0593.50.0093100.5
7B OnlyModel=Qwen2.5-Math-Instruct-1.5B2026.0585.20—