Reasoning Model
A model trained to think out loud before answering, spending extra compute to work through hard problems.
Think of it like
Like a student who scribbles working on scratch paper instead of blurting the first answer that comes to mind.
Example
Ask one a tricky math or coding problem and it’ll generate a long internal chain of steps before giving you the final line.
How it actually works
These models are trained — often with reinforcement learning — to produce extended reasoning traces, then land on an answer. The longer they “think” at inference time, the better they tend to do on hard tasks, which is a real trade of latency and cost for accuracy. It’s a shift from spending all your compute on training to spending some at answer time.
For product teams
Better on genuinely hard tasks, but slower and pricier per query — pick where it earns its keep.
For engineers
A model post-trained to allocate test-time compute to explicit multi-step reasoning before answering.
Related
- Chain of thought — The reasoning trace it produces.
- Test-Time Compute — The compute-at-answer-time idea it embodies.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome