OpenAI has launched two new models, GPT-5.4 mini and GPT-5.4 nano, following the release of GPT-5.4 earlier this month, which was positioned mainly for professional use cases such as programming and data analysis.
GPT-5.4 mini is now accessible to Free and Go users in ChatGPT. Users can select the model by choosing the “Thinking” option from the interface menu.
For paid users, GPT-5.4 mini serves as a fallback option when usage limits for the full GPT-5.4 model are reached.
OpenAI said GPT-5.4 mini delivers improved performance compared to GPT-5.0 mini in several areas, including reasoning, multimodal understanding, and tool usage. The model can better process non-text inputs such as images and audio and has improved capability in tasks like web search.
The company added that GPT-5.4 mini operates at more than twice the speed of its predecessor.
OpenAI is also introducing GPT-5.4 nano, which is designed for tasks that prioritize speed and cost efficiency, such as data classification and information extraction.
Unlike GPT-5.4 mini, the nano model is not available within ChatGPT. Instead, it is offered exclusively through OpenAI’s API for developers.
OpenAI said developers can use larger models to assign tasks to AI agents powered by GPT-5.4 nano. The model is priced starting at $0.20 per million input tokens.