SECAI Core · phase 2 of 15
Small language model
A language model with a relatively low number of parameters and computational requirements, designed to run efficiently on limited hardware (such as edge devices or small servers) while performing focused language tasks
The Explain card
- Plain English
- A small language model (SLM) has far fewer parameters than an LLM, so it runs on modest hardware like phones, laptops or edge devices, handling focused language tasks.
- Example
- An on-device SLM classifies incoming SMS messages as phishing or safe without the text ever leaving the phone, so no cloud service sees the messages.
- Why it matters
- SLMs keep data local, which helps privacy, but they live on devices attackers can physically reach. Extracting the weights or tampering with the model file becomes a realistic threat.
- Hook
- Small enough to fit in your pocket, and in an attacker's.
Where it sits in the deck
Phase 2: How Models Are Built: Architectures and Learning Mechanics
With the paradigms named, zoom in on the architectural building blocks that turn data into trained artifacts.
