Tech words, in plain English.

Large language model: the AI behind chatbots like ChatGPT, Gemini and Claude. It predicts text one piece at a time.

Bigger or newer LLMs usually write, code and reason better, but cost more to run.

A small chunk of text, about 3/4 of an English word. AI prices and limits are counted in tokens.

1 million tokens is roughly 750,000 words, or about 10 novels.

How much text an AI can read and keep in mind at once, measured in tokens.

A bigger window means you can paste whole reports, codebases or books in one go.

The model's trained files are public, so anyone can download it and run it on their own hardware.

You can run it privately and for free, if your computer is strong enough.

The numbers a model learns during training. More parameters often means smarter, but bigger and slower.

A 7B-14B model fits on a normal gaming GPU. 100B+ needs serious hardware.

An AI that takes actions for you, like browsing, filling forms or running code, not just answering.

Agents save time on boring multi-step tasks, but still need you to check their work.

An AI that understands more than text: images, audio, video or files.

You can show it a photo or a screen instead of describing it.

An AI that 'thinks' step by step before answering. Slower, but better at math, code and logic.

Use it for hard problems. For quick questions, a fast model is enough.

When an AI states something false with full confidence.

Always check facts, numbers and links an AI gives you.

A standard test used to compare models or chips with the same tasks.

Useful, but companies pick the benchmarks that make them look best.

A way for apps and developers to use a service, like an AI model, from their own code.

API prices decide how cheap AI features become inside the apps you use.

AI that runs on your phone or laptop, not in the cloud.

Faster, works offline, and your data stays on your device.

How many times per second the screen updates. 120Hz feels smoother than 60Hz.

Once you use 120Hz, 60Hz feels slow. Most good phones have it now.

A measure of screen brightness.

Above ~2,000 nits, the screen stays readable in strong sunlight.

A zoom camera that uses a folded lens to get real optical zoom in a thin phone.

Sharper zoomed photos than cropping the main camera.

Battery size. Bigger is more, but the chip and screen decide real battery life.

Compare real battery tests, not just mAh.

A newer battery type that stores more energy in the same space.

Phones with it often have 6,000+ mAh without getting thicker.

A digital SIM built into the phone. No plastic card needed.

Easy to switch carriers, but check your carrier supports it.

How many years the maker promises new software and security fixes.

7 years of updates means the phone stays safe and useful much longer.

Memory on the graphics card. It holds game textures and AI models while they run.

8GB is tight for new games at high settings. Local AI wants 16GB or more.

The newest type of graphics memory. Faster than GDDR6.

Faster memory helps at high resolutions, but the amount (GB) matters more for most people.

NVIDIA's AI upscaling: the game renders at a lower resolution and AI sharpens it. Only on RTX cards.

More frames per second with almost the same image quality.

AMD's upscaling, like DLSS. The newest version works best on Radeon RX 9000 cards.

Free extra performance in supported games.

AI creates extra frames between real ones, so the game looks smoother.

Feels great above ~60 fps. Below that, it adds input lag.

A way to draw light, shadows and reflections the way they behave in real life.

Looks better, but costs a lot of performance.

How much power a chip or graphics card draws at full load, in watts.

Higher power needs a bigger power supply and more cooling.

The maker's suggested price. Store prices can be higher or lower.

Check real prices before buying. Launch week is often the worst time.

NVIDIA's software for running AI and pro apps on its graphics cards.

Most AI tools work best, or only, on NVIDIA cards because of CUDA.

Neural processing unit: a part of the chip built only for AI tasks, using little power.

It runs on-device AI features without draining your battery.

Trillions of operations per second: a rough measure of how much AI work a chip can do.

Higher is better, but real speed also depends on memory and software.

System on a chip: one chip that holds the CPU, graphics, AI unit and more. Every phone uses one.

The SoC decides most of a phone's speed and battery life.

How a chip is manufactured. Smaller numbers (3nm, 2nm) usually mean faster and more efficient chips.

The number is now more marketing than measurement. Compare real tests instead.

High bandwidth memory: very fast, very expensive memory used in AI data-center chips.

HBM demand pulls memory makers away from normal RAM, which can raise PC and phone prices.