What are Small Language Models?
SLMs are AI models with fewer parameters (1-7 billion) compared to LLMs (70-400+ billion). Examples: Phi-4, Gemma 3, Llama 3.2. Despite smaller size, after fine-tuning they achieve competitive quality in narrow specializations.
SLM vs LLM — when to use which?
SLM: repetitive tasks, classification, data extraction, RAG, query triage. LLM: complex reasoning, long text generation, tasks requiring broad general knowledge. In multi-tier routing, SLMs handle 60-80% of queries while LLMs take the rest.
Enterprise advantages
SLMs run on company servers without sending data to the cloud (privacy + GDPR). Latency is milliseconds instead of seconds. Cost per query near zero. Ideal for industries with strict data requirements: finance, healthcare, public sector.