OpenAI drops GPT-5.4 mini, nano models

The launch of GPT-5.4 mini cements OpenAI's strategy to capture the "agentic" background task market by offering reasoning at scale. An SWE-bench Pro score of 54.38% proves small models can handle complex, multi-file software engineering tasks previously reserved for flagships, while competitive pricing at $0.75 per million tokens undercuts alternatives for high-volume workflows. The new GPT-5.4 nano tier and specialized hardware integrations further enable sub-cent background tasks and ultra-fast inference for responsive developer experiences.