OpenAI announced two new models - GPT-5.4 mini and GPT-5.4 nano, designed specifically for scenarios where response speed is crucial. Developers emphasize that in such cases, priority is shifted from absolute power to instant response and stable integration with instruments. The new products are aimed at applications in fast AI assistants, auxiliary sub-agents, interface analysis systems and multimodal applications for rapid image processing.

The GPT-5.4 mini shows significant progress compared to its predecessor, the GPT-5 mini, in areas such as code generation, logic inference and multimodal analysis, with processing speed more than doubled. In a number of test indicators, it came close to the older model GPT-5.4. The greatest efficiency of the new development is in programming, which requires quick iterations: address editing, navigation through the code base, creating interfaces and debugging.

New models are already available to users: GPT-5.4 mini - via API, Codex and ChatGPT, GPT-5.4 nano - exclusively via API. The mini price is $0.75 per million input tokens and $4.5 per million output, while nano offers lower rates - $0.2 and $1.25 respectively.