What Is an AI Model?

A model is a trained system that does one kind of job. It was shown an enormous number of examples until it became good at predicting what should come next: the next word in a sentence, the next pixel in an image, and what it learned is stored as a large set of numbers called weights.

That is the whole idea. Everything else is consequence.

Three consequences worth knowing

Every model is specialized. A model trained on language is good at language and useless at images. One trained on images knows nothing about your gemstone inventory. There is no single model behind RhinoArtisan’s AI features, because no single model could do all those jobs.

Models are replaced constantly. The strongest model in any category today is usually not the one from six months ago. This is why RhinoArtisan lets you choose your model from a list rather than hardcoding one: when something better appears, you switch to it without waiting for a RhinoArtisan update.

A model does not know things, it predicts them. It has no database to check against and no way to tell a correct answer from a plausible one. It produces the most likely continuation of your prompt. Most of the time that lands on the truth. Sometimes it lands on something fluent and wrong, which is why nothing an AI tells you about a piece replaces the manufacturing checks that measure your actual geometry.

Open weights and closed weights

Models split into two families, and the difference decides almost everything about how you get access to them.

Open-weight models (Llama, Qwen, Mistral and many others) are published. You can download the weights, run them on your own hardware, and pay nothing but electricity. They are free and genuinely useful, and they are noticeably behind the top tier.

Closed-weight models are never published. The weights stay with the lab that trained them, and the only way to use one is to send a request to their servers. Every model at the frontier of what AI can do belongs to this family, which is the subject of the next page.