Conditionally licensed reasoning models

34 in the catalogue today. Every one with what it costs, who sells it and where it stands.

Models that plan before they answer, spending extra tokens on working the problem through. They win on mathematics, hard code and anything with several steps, and they are measured on boards like GPQA, AIME and ARC-AGI. The catch is what the thinking costs: reasoning is billed as output tokens, so the same model can cost several times more per answer at a high effort than a low one, and for a question that needed one turn you have paid for a monologue.

Weights you can download, under a licence that asks for something in return: an acceptable-use policy, a naming requirement, a revenue ceiling above which you must ask, a restriction on training other models. The Llama and Gemma families are the familiar cases. Read the actual licence before you build on one of these — the condition is usually easy to meet and occasionally fatal, and which it is depends on your product rather than on the model.

Wider

Conditionally licensed models Reasoning models