GPT, Claude and friends: language models in plain English

If you've decided to finally get into AI, you've probably noticed everyone throwing around names like GPT, Claude and Gemini, plus words like “models”, as if you're supposed to already know what they mean. It can feel like alphabet soup, and it's easy to assume there's some obvious thing everyone understands except you.

There isn't. These are just product names and a bit of jargon, and once someone explains them in plain English they stop being scary. So let's do exactly that, slowly, with no assumed knowledge. By the end you'll know what these things are and, just as usefully, what you're allowed to ignore.

What a language model actually is

Think of a chat AI like a car. The chat window you type into is the dashboard, but the engine under the bonnet, the part that actually produces the words, is the large language model, often shortened to LLM. When people say “model”, they usually mean this engine.

How did it get so good with words? In simple terms, it learned by reading. An enormous amount of text, from books to websites, was fed into it, a bit like everything a person might read growing up. We call that the training data. By seeing so much language, it got very good at predicting what words tend to come next, which is how it holds a sensible conversation with you.

So what are GPT, Claude and Gemini?

Here's the part that trips people up. GPT, Claude and Gemini are not different kinds of thing. They're just different families of these engines, each built by a different company. It's a bit like Ford, Toyota and Volkswagen all making cars: same basic idea, different makers, slightly different feel.

• GPT is the family made by a company called OpenAI, and it's what powers the well-known ChatGPT.
• Claude is the family made by Anthropic.
• Gemini is the family made by Google.
• There are others too, but these are the names you'll bump into most often.

You don't need to memorise who makes what. The useful thing to know is simply that they're rival versions of the same idea, and any of them will happily answer your questions.

What do names like Haiku, Sonnet and Opus mean?

Within one family you'll often see a second name, like Haiku, Sonnet or Opus. These aren't different products so much as different sizes of the same engine. Think of coffee cups: small, medium and large. They all pour the same coffee, just more or less of it.

• Smaller models are faster and cheaper, and great for quick, everyday questions.
• Bigger models are a little slower but can handle trickier, more thoughtful tasks.
• For most things you'll do as a beginner, even the smaller ones are more than enough.

Three quirks worth knowing

A few things about these engines surprise newcomers, so it's worth naming them gently. First, an AI has a knowledge cut-off. Because it learned from training data gathered up to a certain date, it often won't know about very recent events, a bit like a brilliant book that was printed last year.

Second, in a single conversation it only has so much short-term memory, called the context window. Think of it like a desk with limited space: if a chat goes on very long, the earliest things can slide off the edge and be forgotten. And third, it can sometimes be confidently wrong, stating something untrue as if it were fact. That's called a hallucination, and it's simply a good reason to double-check anything important.

What now?

That's genuinely the core of it. A language model is the word-producing engine; GPT, Claude and Gemini are just different makers; and names like Haiku, Sonnet and Opus are sizes within a family. Everything else you can pick up gently as you go.

If you'd like to keep going at this calm pace, Bliks Academy is built for exactly this, and you can always practise with Sophia, our friendly sandbox, with no pressure. And any time a word here felt new, the glossary is there to explain it in the same plain English.