AI agent reasoning level

An AI agent reasoning level is how much the AI model thinks before it replies. In Inrō the Reasoning menu appears once you save your own OpenAI API key; without one, Inrō chooses the reasoning level for you. A lower level answers faster and uses fewer tokens. A higher one thinks more and costs more.

What levels can you pick?

The Reasoning menu sits next to the Model menu in the OpenAI API Key section, under Profile, Subscription & add-ons. The levels, from cheapest to most thorough:

1. Minimal
2. Low
3. Medium
4. High
5. Extra High

Default (low) runs the agent at Low. The menu lists only the levels your chosen model accepts. Switch to a model that refuses your level and Inrō moves it to the nearest one the model takes.

How does each level change speed and cost?

OpenAI puts the trade-off plainly in its reasoning guide (checked October 2026): lower effort favours speed and lower token usage, while higher effort lets the model think more completely. You never see reasoning tokens, but OpenAI bills them as output tokens.

• Minimal: the least thinking before a reply, the quickest and cheapest step. Only some models accept it.
• Low: a little more thinking. This is the default.
• Medium: more thinking time and more output tokens than Low.
• High: more again, so slower and dearer replies.
• Extra High: the most thinking, so the slowest and most expensive step per reply. Only some models accept it.

Once you move the model or the level off its default, the app warns you: "Using a higher model or reasoning level may increase reply latency and costs."

Who pays for the extra reasoning?

You do, through OpenAI. The menu only shows with your own key, and the agent's replies run on it. OpenAI bills the tokens to your account, and Inrō charges no AI credits for calls on your own key.

What does the reasoning level not change?

It changes how hard the model thinks, not what it knows. Facts about your products and prices come from the agent knowledge base. What the agent says when it has no answer is set by its fallback.

How the agent sounds comes from its style rules, which the brand voice generator for an Instagram DM agent drafts. The Inrō API for developers returns the level with the agent's configuration; you change it in the app.

See the Inrō AI agent for Instagram DMs and the Inrō automation glossary.