Reasoning model
A model designed to work through a problem in multiple internal steps before answering, trading additional inference time for higher accuracy on complex tasks.
In more detail
A reasoning model is tuned to break a hard problem into intermediate steps before producing a final answer, rather than answering in a single pass. On tasks with multiple dependent steps, planning, or careful constraint-checking, this deliberate approach tends to be more accurate than a fast single-shot response.
The trade-off is time and cost. Reasoning takes more inference, so it makes sense to reserve it for the steps that genuinely benefit: planning a multi-step workflow, resolving a tricky exception, or checking work. Simple lookups and classifications do not need it.
Where this shows up at Ceven
Ceven can put a reasoning model on the steps that need it, such as planning how to satisfy a plain-language request or resolving an exception, while lighter steps run on faster models. Wide and deep research also lean on careful multi-step reasoning to produce briefs with citations rather than an unsupported summary.