Glossary · AI
What is
Small Language Model (SLM)?
A compact LLM (typically 1-15B parameters) optimized for specific tasks or constrained environments.
By Anish· Founder · Vedwix
·Definition
Small Language Models trade general capability for cost and latency. For narrow tasks (classification, structured extraction, simple Q&A), a fine-tuned 7B model can outperform a frontier model at 1/30th the cost. SLMs are crucial for high-volume use cases and on-device AI.
Example
A fine-tuned Phi-3 (3.8B params) for invoice extraction, running 100k requests per day at $0.0001 each.
How Vedwix uses Small Language Model (SLM) in client work
Most production AI features end up as fine-tuned SLMs. We start with frontier models and graduate down.
Building with Small Language Model (SLM)?
We ship this.
If you're building with Small Language Model (SLM) in production, we can help — from architecture review to full implementation.
Brief us