Lesson 4 of 6
How Big Is an AI? Models, Training Data and VRAM
ChatGPT doesn't fit on your phone. Explore why AI models need gigabytes — and their training data needs data centres.
Learn it
An AI model is essentially a huge file of numbers called parameters. Each parameter is typically stored as 2 bytes.
A small phone-friendly AI model might have 3 billion parameters ≈ 6 GB — too big for most phones to run well, which is why many AI features run in the cloud.
Large models like GPT-4-class systems have hundreds of billions of parameters — hundreds of GB — and need multiple datacentre GPUs just to run.
The training data is vastly bigger than the model: models learn from hundreds of billions of words — many terabytes of text.
Key terms
- Parameter
- One number inside an AI model; billions of them together encode what the model knows.
- Quantisation
- Storing parameters with fewer bits (e.g. 4-bit) to shrink model size.
- VRAM
- A GPU's fast memory; the model must fit inside it to run.
- Training data
- The text/images the model learned from — far bigger than the model itself.
Try it
Calculate the size of an AI model: 7 billion parameters × 2 bytes each. Fill in the missing value in GB.
parameters = 7_000_000_000
bytes_per_param = 2
size_bytes = parameters * bytes_per_param
size_gb = size_bytes / 1_000_000_000
print(size_gb) # Output: ______Challenge
Phone AI planner: your phone has 8 GB RAM, but 5 GB is already used by the OS and apps. Can you run a 3B-parameter model at 2 bytes per parameter? What about a quantised 4-bit version of a 7B model? Show your working.
Pick whichever way suits you — every mode earns the same bonus XP.
Write at least 40 more characters to submit.
Mark your own work
Guided walkthrough — 0/3 clues revealed
- Clue 1 locked — reveal it only if you get stuck.
- Clue 2 locked — reveal it only if you get stuck.
- Clue 3 locked — reveal it only if you get stuck.
Each clue costs 3 XP (never below 15 XP). You'd earn 30 XP right now.