OpenAI Releases GPT-5.4 Mini and Nano: Fast, Cheap, and Built for AI Agents

What happened

OpenAI released GPT-5.4 mini and GPT-5.4 nano on 17 March 2026, describing them as the "most capable small models yet." They are smaller, faster, and cheaper versions of GPT-5.4, designed for tasks where speed and cost matter more than raw power.

GPT-5.4 mini runs more than twice as fast as the previous GPT-5 mini and performs close to the full GPT-5.4 model on several benchmarks, including coding and reasoning tasks. GPT-5.4 nano is the smallest and cheapest option, aimed at classification, data extraction, and supporting tasks in automated workflows.

Both models are built with AI agents in mind. They are designed to handle the kind of repetitive, fast-turnaround work that larger models would be overkill for, such as reviewing files, navigating a codebase, or running parallel subtasks in the background. ChatGPT users can access GPT-5.4 mini today.

What this means for your business

If you are using AI tools that run behind the scenes in your software or workflows, like coding assistants, customer service bots, or document processors, these models make that kind of automation significantly cheaper and faster to run.

For business owners building with AI, or evaluating tools that use AI under the hood, this is a sign that the cost of running capable AI is continuing to fall. What required expensive compute six months ago can now be done at a fraction of the price. That matters if you are thinking about adding AI features to your own products or services.