10-18-2026, 10:18 PM
Can you fill in a bit more context? What model, what runtime, and roughly what context size?
small models can still be good
|
on-device inference worth the effort: what are you seeing?
|
|
« Next Oldest | Next Newest »
|