Fine-Tuning LLMs · Lesson 1
Introduction to SLMs
Small Language Models (roughly 0.5–14B parameters) such as Phi, Gemma, Llama small variants, Qwen and Mistral 7B can run on a single GPU, a laptop or even a phone.
They are cheaper and faster, keep data on-premise, and when fine-tuned on a narrow task they can match much larger general models.
They have less broad knowledge and weaker complex reasoning, so pair them with retrieval or route hard cases to larger models.
Key points
- Run on modest hardware
- Lower cost, latency and privacy risk
- Strong on narrow tasks after fine-tuning
- Weaker general reasoning