K-Culture GlossaryㅍWhere everyone starts
Parameter
The number of adjustable dials inside a model. The "B" in "70B model" refers to this — generally, more means smarter but more expensive.
In plain words
A parameter is one of the "adjustable dials" inside an AI model. The knowledge and patterns a model learns get stored as the values of these dials. More dials mean the model can capture finer-grained patterns, so the parameter count is used as a number that represents a model's "weight class."
The "7B" or "70B" in article titles refers to exactly this. B stands for Billion — a 7B model has 7 billion dials. Think of it like boxing's lightweight versus heavyweight divisions. A bigger weight class is usually smarter, but running it requires better equipment (GPUs) and more electricity.
But "bigger is always better" is old news from a few years back. These days, the interesting stories go the other way — small models trained well keep beating bigger ones, and compact models that run on phones (on-device) have become their own genre.
How it shows up in the news
"With just 3B parameters, it matched 70B-class performance" — this is one of the most popular headline formats in this field: a lightweight model taking on one in a higher weight class.
Tokens vs. Parameters: What's the Difference
Is it the count of numbers inside the model?, Is it the amount of text going in and out?
Both are numbers used to describe scale, so when an article mentions something like '70 billion,' it's easy to lose track of which one is being discussed. One is the amount of text you send and receive, the other is how big the model itself is. If it grows with how much you use it, it's the former; if it's fixed when the model is built and never changes, it's the latter.
| Aspect | Parameter | Token |
|---|---|---|
| What it counts | The number of adjustable knobs inside the model | Units counted by splitting up text exchanged |
| Does it change? | Fixed when the model is built and never changes | Grows every time you use it |
| Relation to cost | The bigger it is, the more expensive it is to run | Pricing is based on this |
| In articles | The B in "70B model" refers to this | "1 million token context," "price per token" |
| Analogy | The number of brain cells someone has | Minutes spent on a phone call |
Rule of thumbIf it grows with how much you use it, it's tokens; if it's fixed when the model is built and never changes, it's parameters.
See also
Stories using this term
No story has used this term yet. New ones attach here automatically.