OpenAI’s GPT-5.4 mini and nano models cut costs and latency while staying close to flagship performance, giving developers ...