Parameters (7B, 70B)
Parameters are a model's learned weights, counted in billions — the B in 7B. More parameters mean more memory, and only sometimes better answers.
Every capability a model has lives in its parameters — billions of numbers tuned during training. The count is the closest thing AI has to an engine-size figure, and like engine size it is often read as more than it is.
Two things it genuinely determines: memory and compute. Weight size scales directly with the count times the storage precision, which is why quantization matters so much locally. What it does not guarantee is quality: training data, technique and architecture move results at least as much, and newer small models regularly beat older large ones.
For local use, the practical translation is simple: the parameter count sets the class of hardware you need, the measured file size settles whether it fits, and neither number tells you whether the model is good at your task — that part you judge by using it.
Where you see it on the radar
We print measured file sizes next to parameter counts — the download is what has to fit in memory, not the marketing number.
See sizes as measured, not assumed →Related terms