Model architecture
Reasoning model
What is reasoning model?
A reasoning model has been trained to produce explicit intermediate steps before its final answer. The visible working improves accuracy on multi-step problems and costs proportionally more tokens per response.
The improvement is real on maths, code and multi-hop questions, and largely absent on simple retrieval or classification — where it just makes responses longer and slower.
A reasoning-distilled base is a strong starting point when the target task needs step-by-step behaviour, because the behaviour is present before fine-tuning begins.
Related terms
Start with the free tier
A magic link creates your account, your tenant and your first API key. No card until you ask for compute.