OpenAI Just Revealed Cheaper Versions of Its Flagship Model. Here’s How to Use Them |
OpenAI Just Revealed Cheaper Versions of Its Flagship Model. Here’s How to Use Them
OpenAI’s GPT-5.4 mini and nano are here, and they’re designed to help you get stuff done on a budget.
BY BEN SHERRY, STAFF REPORTER @BENLUCASSHERRY
Illustration: Inc.; Photos: Courtesy OpenAI
OpenAI has introduced GPT-5.4 mini and GPT-5.4 nano, two small, cost-efficient versions of its flagship AI model.
In a press release, OpenAI said that 5.4 nano and mini are “our most capable small models yet,” coming close to matching the default GPT-5.4 model’s abilities at coding and agentically operating software. But they come at a much cheaper price. The new models are also significantly faster than the flagship model and previous small models, making them particularly useful for collaborative vibe coding work.
OpenAI typically releases models in four sizes; a large pro version, a middle-of-the-road standard version, a small mini version, and an even smaller nano version. Smaller models are faster and cheaper, but not as capable as their larger siblings.
Both models excel in “workloads where latency directly shapes the product experience,” according to OpenAI, such as coding assistants that can operate and make changes in real time and computer-use agents that can handle data entry.
How Canva Became the Power Player in the AI Design Wars
For ChatGPT users, OpenAI recommends using GPT-5.4 mini for small tasks that are very well-defined, like editing pages of text that need to keep to a very specific format. On Codex, OpenAI’s agentic coding tool, the company says that 5.4 mini is well-suited for being used as a sub-agent, in which a larger “manager” model handles planning, coordination, and judgment of a task, while a collection of smaller “worker” models simultaneously attack individual aspects of that task, like searching a codebase or processing documentation.
“Instead of using one model for everything, developers can compose systems where larger models decide what to do and smaller models execute quickly at scale,” OpenAI wrote. “GPT-5.4 mini is our strongest mini model yet for that style of workflow.”
On the other hand, OpenAI says that GPT-5.4 nano should only be used for tasks where speed and cost-efficiency are highly important. If you have a very simple task that needs to be replicated hundreds or even thousands of times per day at near-instant speed, this might be the model for you.