Are Small Language Models (SLMs) better than LLMs for building AI agents? by Agentic Workforce Key Takeaways Power without bloat : Models as small as 2.7B parameters (e.g., Microsoft’s Phi-2) now match or exceed the reasoning and code generation of 30B+ models — while running up to 15× faster .... AI Architecture AI Strategy Small Language Models Sustainable AI Nov 11, 2025