The GPU model is the best clue
Chrome, Edge and similar browsers report the exact GPU, such as Apple M2, RTX 4070 or Adreno 740. Each model has a clear release year and market position, so it maps straight to a year and class.
Without any permission, a web page can read your GPU model, system version and screen size. This page uses them to guess what year your device is from and which class it's in.
To draw 3D graphics a page can read your GPU model; to lay out content it can read the screen size and pixel ratio; the system version and CPU core count are readable too. None of it needs permission, and together it reveals your device's rough age and class.
Chrome, Edge and similar browsers report the exact GPU, such as Apple M2, RTX 4070 or Adreno 740. Each model has a clear release year and market position, so it maps straight to a year and class.
Browsers on iPhone just say 'Apple GPU'. But each iPhone generation has its own combination of screen size and pixel ratio, which narrows it down to a few models, and the system version rules out the oldest ones.
The year is the model's release year; you may have bought it later, and the same GPU appears in machines at different prices. Safari and Firefox hide or blur the GPU model, leaving only a lower bound from the system version. Just for fun.
No. Everything is worked out in your browser and is gone when you reload.
Safari, Firefox and some privacy browsers hide the real GPU model and report only a generic name. Opening the page in Chrome or Edge usually gives a more specific result.
Browsers deliberately round memory and cap it at 8 GB so sites can't use the exact amount to identify your device.
Ad systems use it to guess your spending power, so a new flagship and an old budget machine may see different ads. Anti-fraud systems also check that device details are consistent, such as a phone claiming to be an iPhone but reporting a desktop GPU.
The year is the model's release year and the class reflects its position at launch, which may differ from when you bought it or its exact configuration. If the GPU was identified wrongly, let us know through the feedback page.