What is an AI model, and why do new ones come out every week?

You use AI through apps: ChatGPT, Claude, Gemini, Copilot. Underneath each app is a model, the trained system that actually reads your request and writes the answer. When a company announces a “new model”, the app usually stays the same and gets smarter (or faster, or cheaper) underneath.

The main kinds of models

Most releases fall into a few groups. Our model tracker labels every release with one of these.

• Flagship models. The biggest and most capable models a company offers. Best for hard reasoning, long documents and complex coding. Often slower and pricier.
• Fast and cheap models. Smaller versions tuned for speed and cost. Great for everyday tasks, summaries and anything run at scale.
• Open-weight models. Models whose weights you can download and run yourself. They matter to developers and businesses that need privacy or control.
• Image, video and voice models. Specialists that create pictures, video clips or speech.
• Coding and specialist models. Trained for one domain, such as writing code or transcribing audio.

How to tell if a new model matters to you

Ask three questions when you see a launch headline:

1. Can I use it today? Many announcements are previews, waitlists or developer-only. Our tracker shows the status of each release.
2. Is it in an app I already use? If yes, you may get it automatically.
3. Is it better at my task? Benchmarks measure general skill. What matters is whether it summarizes your reports or writes your emails better. That’s what our use-case guides test.

Why the names are so confusing

Companies release many versions: big and small, fast and slow, numbered updates in between. A good rule: the name tells you who made it and roughly how new it is; the size label (such as “Flash”, “Mini” or “Haiku”) tells you whether it is tuned for speed.

Bottom line

You don’t need to follow every launch. Check whether a new model is available to you and whether it improves a task you care about. Our new model tracker does the first part for you, and our use-case guides do the second.