Y
Hacker News
new
|
ask
|
show
|
jobs
by
agsqwe
369 days ago
thinking models produce a lot of internal output tokens making them more expensive than non-reasoning models for similar prompt and visible output lengths