Reasoning model
A reasoning model works through a problem before answering, rather than answering straight away.
It spends more output tokens and more time, and does much better on anything with steps.
On a simple rewrite it is slower for no gain. The skill is knowing which kind of job you have.
Was this page helpful?