Let's talk about the next generation of AI: reasoning models, like the OpenAI o1 and 5 series.

These models introduce a brand-new parameter called "Reasoning effort," which allows you to control how hard the AI thinks before answering.

The usability challenge for us is that this parameter only works with specific models. If you try to send a "Reasoning effort" setting to a standard model like GPT-4.1, the OpenAI API will return an error and reject your prompt.

This is where the "Auto-set" feature comes in.

When you leave the reasoning effort on "Auto-set," even if you select a reasoning model, the correct defaults are applied. But if you switch to a standard model, it intelligently "hides" that parameter so it isn't sent to the API.

This means you don't have to memorize which models support reasoning and which don't. You can switch profiles freely without ever triggering an API compatibility error.

But of course, power users can customize to their heart’s content.

Also, don’t forget how much more expensive reasoning models can be than regular ones. Watch your OpenAI account charges if you start using reasoning models extensively and on high effort settings.