# Llama 3.2 1B / 3B — Edge & Mobile FT: Tied Embeddings + Distillation + GGUF Q4

> Source: https://sukruyusufkaya.com/learn/fine-tuning-cookbook/ftc-llama-3.2-1b-3b-edge-mobile-ft
> Updated: 2026-08-21T13:26:14.369Z
> Category: Fine-Tuning Cookbook (Model-by-Model)
> Module: Part III — Small Open Models (1B–8B)
**TLDR:** Llama 3.2 1B/3B — Llama 3.1 8B'den distilled. Tied embeddings (input/output emb shared), edge cihazlarda inference. RTX 4090'da full FT mümkün (1B=2GB, 3B=6GB W). GGUF Q4_K_M quant ile iPhone/Pixel'de 8-15 tok/s. TR-MMLU sayıları ve dataset stratejileri.