OpenAI drops GPT-5.4 mini and nano: faster, cheaper AI

OpenAI just released GPT-5.4 mini and nano, two new small models that pack a serious punch. Think of them as the efficient workhorses of the GPT-5.4 family. The mini version is over 2x faster than its predecessor, GPT-5 mini, and nearly matches the big GPT-5.4 model on coding and reasoning benchmarks like SWE-Bench Pro. The nano is the smallest, cheapest option yet, perfect for simple tasks like classification or data extraction. Both models are designed for scenarios where speed matters—coding assistants, subagents, and apps that need to think fast about images or screenshots. Early testers report that mini actually outperforms the larger GPT-5.4 on some citation and output tasks, at a fraction of the cost. The bigger strategy here is about composing systems: let a large model plan and delegate, while smaller models execute quickly in parallel. Why it matters: This release signals that the AI arms race is shifting from pure size to smart orchestration, where combining different-sized models can beat a single monolithic one on both performance and cost.