Generative AI · Fast-moving · Intermediate

Small Language Model

Also known as: SLM

A compact language model, typically under roughly ten billion parameters, designed for cheap, fast or on-device use.

What Small Language Model is

Small models trade broad world knowledge for speed, privacy and cost. With good data and fine-tuning they can match much larger models on narrow tasks.

How it works

They are produced through efficient pretraining on high-quality curated data, distillation from larger teachers, and quantisation for deployment.

Why it matters

Most production workloads are narrow. Routing those to a small model and reserving a frontier model for hard cases is now standard cost engineering.

Common uses

→On-device assistants
→Classification and extraction at volume
→Offline and privacy-sensitive deployments
→Edge and IoT

Strengths

✓Cheap and fast
✓Runs locally without sending data out
✓Fine-tunes affordably

Watch for

✓Weaker general reasoning
✓Smaller effective knowledge

Continue exploring

Quantization Model Distillation Edge AI Open-Weight Model