Company, product, or model?

You’ll hear “ChatGPT,” “GPT,” and “OpenAI” used as if they were one thing. Same with “Claude” and “Anthropic,” or “Gemini” and “Google.” They aren’t, and the confusion isn’t your fault. Some of these names really are used for two things at once.

---

Three layers

The company builds and sells the technology. OpenAI, Anthropic, Google, Microsoft.

The product is what you open and use: a website, a phone app, or a feature inside another program. It’s the part with buttons, settings, a sign-in, and a price.

The model is the AI underneath. An AI model is the trained system that actually generates the responses; a product wraps it in something you can use. You rarely deal with the model directly. You reach it through a product.

A rough comparison: the company is the carmaker, the product is the car you drive, and the model is the engine. The same engine can go into several cars, and one car can offer more than one engine.

---

The well-known examples

• OpenAI is the company. ChatGPT is the product. The models underneath are the GPT family.
• Anthropic is the company. Claude is the name of the app and of the model family.
• Google is the company. Gemini is the name of the app and of the model family.
• Microsoft is the company. Copilot is the product. It runs on models from OpenAI, among others.

Two things make this harder than it looks.

Some names do double duty. “Claude” can mean Anthropic’s app or the models inside it. “Gemini” can mean Google’s app or its models. You have to tell from context, and the companies’ own pages switch between the two.

Model names change constantly. New versions come out several times a year, with numbers and suffixes attached. This page deliberately doesn’t list them, because any list would be out of date within months.

---

Why the difference matters to you

The same model can behave differently in different products. A company that builds a product on top of a model adds its own instructions, limits, and features. Two apps using the same model can give noticeably different answers.

The same product can change models under you. An app you’ve used for a year may be running a different model today than when you started. If it suddenly writes differently, that’s one possible reason.

Plans can differ. Free and paid plans for the same product can differ in which models you can pick and how much you can use them. Check the product’s own pricing page rather than a headline.

Headlines usually mean the model. “New model released” is about the engine. Whether your app has it yet, and on your plan, is a separate question.

---

Something you can try

Open the AI app you use, or the one you’d try first, and look for a model name. It’s often in a menu near the top of the chat or in settings. Some apps show it. Some don’t.

Then ask it:

Which company makes you, and what is the difference between you as
an app and the model you're running on?

Compare its answer with what the app’s own settings or help pages say. If they disagree, trust the settings page.

---

Something important to know

A new model name isn’t proof it’s better at your task. Companies announce new models with test scores. Those tests aren’t your work. The test that tells you whether a tool is good for your task is trying your task on it and checking the result.

---

If it doesn’t match what you see

Your app doesn’t show a model name at all. Some don’t, and that’s fine. You can use a tool well without knowing which model is underneath.

The names on this page have changed. Companies rename and merge products. The three layers stay the same even when the names don’t.

Someone at work says “we use GPT” or “we use Claude.” Ask which product, and under what kind of account. A company account and a personal account for the same product can have different rules about your data, which the plain answers cover.