Phi-3 mini 3.8B is among the smallest credible LLMs for agentic use in 2026. The 4-bit quantised version runs at ~6 tok/s on a Pi 5. Quality trails larger models meaningfully on complex tasks but is genuinely useful for narrow workflows.
Related terms
Found a definition that's wrong, dated or could be sharper? Email us — we update with attribution unless you'd rather we didn't.