Very roughly, looks like o3-mini is about as good as o1 but cheaper and faster (close to 50/50 in heads up human preferences). The main goal is probably to addr

Very roughly, looks like o3-mini is about as good as o1 but cheaper and faster (close to 50/50 in heads up human preferences). The main goal is probably to address the ballooning inference costs that OpenAI is racking up.