Artificial intelligence is software that performs tasks we normally associate with human thinking, like recognising what's in a photo, understanding a sentence, or writing a paragraph of text.
An AI model is the trained pattern-matching engine inside that software. It's the part that has learned from examples, and it's what people mean when they say a tool "runs on" a particular model.
The tool itself — the app you open, the buttons you click — is a separate layer sitting on top. If you only remember one thing, remember this: the model is the engine, the tool is the dashboard, and the two change on completely different schedules.
What AI actually is
Regular software follows instructions a person wrote. A calculator adds because someone coded the rules of addition. AI works differently: instead of being told the rules, it's shown thousands or millions of examples and adjusts itself until it gets good at spotting the pattern.
That adjustment process is called training, and the frozen result of training is the model. When you use the tool, it isn't learning from you in that moment — it's running the pattern it already learned. That moment is called inference, and it's usually much cheaper and faster than training.
This is why a tool can feel smart on day one and still be wrong on day two: the model didn't get worse, it just met a case its training didn't cover well.
How a model is made, and why version numbers matter
Training happens in rounds. Engineers gather examples, run the training, test the result, then tweak and run it again. Version numbers — the V7, the v27.5, the "2.0" you see in product names — mark points where a team decided the result was meaningfully different from the last one.
Sometimes the change is size (more examples, more capacity). Sometimes it's a new capability, like handling images as well as text. Sometimes it's mostly a fix for a bad habit.
According to our AI tool database, Midjourney's V7 generation added a Draft Mode, a video model, and personalisation features that earlier versions didn't have — which is a good illustration of how a version bump can mean genuinely new abilities rather than a cosmetic rename. Adobe's Photoshop 2026 (v27.5) carries Firefly-powered generative fill and AI editing tools, and the version label there signals which editing features exist in that release.
The practical point for a beginner: a version number is a claim about capability, not a guarantee. Two tools can both say "AI-powered" and be running models of very different quality.
One worked example
Say you use an image tool to turn a rough sketch into a finished illustration, and the results suddenly look flatter than last month. Three things could have changed. The model could have been swapped or retrained.
The interface could have changed — maybe a default setting quietly moved from "high detail" to "fast". Or your own input changed, because you started writing shorter prompts. Here's a decision rule you can actually use: change one thing at a time and keep a note.
Save your exact prompt, run it twice on different days, and compare. If identical inputs now produce different outputs, the model or its settings changed underneath you. If identical inputs still produce identical outputs but the button you press is in a new place, only the interface changed — and that's a much smaller problem. Most people blame the model when the real culprit is a default they never noticed.
Where this breaks down
Knowing that a model exists doesn't tell you whether a tool is any good, and it doesn't tell you what it costs. Pricing and plan structure change often, and the vendor's own page is the only reliable source — our database snapshot records Midjourney's Basic tier at $10 a month and Photoshop's standalone plan at $20.99 a month at verification time, but those figures are a snapshot, not a promise.
The model-versus-tool distinction also stops being useful once you're dealing with a tool that swaps models silently, which many do. And a newer version is not automatically better for your specific job: a version tuned for photorealism may be worse at line art, and a version tuned for speed may lose the fine detail you needed.
The honest answer is that version numbers tell you what changed, not what's right for you.