OpenAI launches GPT-5.4 mini and nano for faster, cost-efficient AI workloads

OpenAI has announced the release of GPT-5.4 mini and nano, its most capable small language models built for speed and efficiency in high-volume AI workloads. Both models inherit key strengths from the larger GPT-5.4, offering notable improvements in performance and utility for developers seeking rapid responses.

Compared to its predecessor, GPT-5 mini, the updated GPT-5.4 mini runs over twice as fast and demonstrates strong results across coding, reasoning, multimodal understanding, and tool use. On evaluation benchmarks such as SWE-Bench Pro and OSWorld-Verified, GPT-5.4 mini approaches the performance of the more resource-heavy GPT-5.4, making it suitable for demanding yet latency-sensitive applications.

Following this, GPT-5.4 nano emerges as the smallest and most cost-effective GPT-5.4 option available. OpenAI recommends GPT-5.4 nano for scenarios where speed and affordability are critical, such as classification, data extraction, ranking, and subagents supporting larger systems. Both models prioritize tasks like coding assistants, fast completion of supporting subroutines, real-time image reasoning, and rapid screenshot analysis.

Building on these capabilities, GPT-5.4 mini and nano support fast-iteration coding workflows, including codebase navigation, debugging, and front-end generation. The models are also suitable for use in systems that combine different model sizes and excel in interpreting complex user interface screenshots for computer use tasks. GPT-5.4 mini and nano are now accessible via the API, Codex, and ChatGPT.