Model size
Roughly how big a model is, usually counted in parameters — the headline number everyone quotes.
Think of it like
Engine displacement on a car: bigger often means more power, but it’s not the whole story.
Example
A “70B” model has 70 billion parameters; a “7B” has ten times fewer, and runs far cheaper.
How it actually works
Size is usually parameter count, a proxy for capacity. Bigger models tend to be more capable but cost more to train and run, and need more data to earn their size. Recent trends show a well-trained small model can beat a bloated large one — size is a dial, not a destiny.
For product teams
Bigger isn’t automatically better; weigh capability against latency and cost for your task.
For engineers
Parameter count (and by extension memory/FLOPs); one axis of the scaling-law trade-off.
Related
- Parameters — Counted as parameters.
- Frontier model — Trades off in a frontier model.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome