Small language model (SLM)
A compact language model that trades peak reasoning ability for lower latency and cost, well suited to well-scoped tasks like classification, extraction, and routing.
In more detail
Small language models have far fewer parameters than frontier models. They give up some capability on the hardest reasoning tasks in exchange for much lower latency and cost, and they are often good enough for narrowly scoped work: labeling a message, pulling fields out of a document, or picking a branch.
The engineering question is fit, not size. A workflow that sends every step to the largest available model is usually paying too much and waiting too long. Matching each step to the smallest model that reliably handles it is how automation stays fast and affordable at volume.
Where this shows up at Ceven
Ceven matches each workflow step to an appropriately sized model, so a high-volume classification or extraction step does not pay frontier-model latency and cost when a smaller model handles it reliably. You describe the outcome once; Ceven handles the routing underneath so the workflow stays fast and inexpensive at scale.