Parameter count
Short answer
Parameter count is how many learned weights a model has — the 2B, 9B or 12B in its name, meaning billions.
More parameters generally means more capability and always means more memory. It is a rough guide rather than a ranking: on Nightjar's 16-question evaluation, Qwen 3.5 2B scored 16 out of 16 while the much larger Qwen 3.5 9B scored 14.
That result is worth sitting with. For reading a page in front of you, how recently and how well a model was trained matters more than how big it is.
Related terms
Other ways people ask this
- what is parameter count