OpenAI broadens its developer model stack with GPT-5.4 mini and nano

OpenAI Developers said in an X post on March 17, 2026 that GPT-5.4 mini and GPT-5.4 nano are now part of the developer lineup, extending the company's small-model strategy beyond the flagship GPT-5.4 release. The announcement points developers to OpenAI's product note, which describes the pair as the company's most capable small models yet.

The split is straightforward. GPT-5.4 mini is designed to be the fast general-purpose workhorse for coding, multimodal understanding, tool use, and subagent-style workflows. OpenAI says it runs more than 2x faster than GPT-5 mini while improving performance on coding and tool benchmarks such as SWE-Bench Pro, Terminal-Bench 2.0, Toolathlon, and OSWorld-Verified. GPT-5.4 nano sits one step below it as the lowest-cost option for lighter classification, data extraction, ranking, and supporting agent tasks.

That matters because modern AI products increasingly mix models rather than relying on one frontier model for everything. In the OpenAI framing, a larger model can plan or judge results while smaller models handle narrower tasks in parallel. The company explicitly calls out coding assistants, computer-use systems, screenshot interpretation, and real-time multimodal applications as workloads where latency matters enough that a faster small model can be the better product choice.

Availability also reinforces that positioning. OpenAI says GPT-5.4 mini is available in the API, Codex, and ChatGPT, with a 400k context window and support for text and image inputs, tool use, function calling, web search, file search, computer use, and skills. GPT-5.4 nano is API-only and priced below mini for high-volume workloads. For developers building agent systems, the launch is less about one more model name and more about a clearer cost-performance ladder inside the GPT-5.4 family.