Small language model (SLM)
What is Small language model (SLM)?
A language model small enough to run cheaply, sometimes on your own hardware or even a phone. It trades broad general knowledge for lower cost, speed and easier control, and it does narrow, well-defined jobs well. As a rough line, anything under about 30 billion parameters gets called small.
Why it matters
Not every job needs the biggest, most expensive model. A small language model is compact enough to run cheaply, sometimes on your own servers or even a phone, and for a narrow well-defined task it can match a large one at a fraction of the cost. For a business that changes the maths of putting AI into a product. Running a huge model on every request adds up fast; a small one tuned to a single job keeps the unit cost low, and it can run on your own hardware so sensitive data stays in-house. The trade is breadth. An SLM does its one thing well and lacks the general knowledge of a large model.
In practice
A manufacturer runs a small model on the factory floor to read sensor logs and flag faults, with no data leaving the site. A software team uses one for a single classification step inside a larger pipeline where a general model would be overkill. As a rough line, anything under about 30 billion parameters gets called small. The point is fit for purpose, not size for its own sake.
