LocalForge
Back to guide
MODEL / META

Llama 3.2 3B

Long advertised context still increases KV-cache memory; use modest context on small devices.

Primary source
Total parameters3B
Estimated memory~2.8 GBGGUF Q4
Context128KText
LicenseLlama 3.2 Community8 official languages
01 / BEST FOR
  • chat
  • summaries
  • mobile
02 / ESTIMATED FIT
8 GB CPU laptopComfortable16 GB CPU desktopComfortableApple silicon · 16 GBComfortableNVIDIA · 8 GBComfortableNVIDIA · 12 GBComfortable
03 / DEPLOY WITH
OllamaEasyollama run llama3.2:3b
llama.cppMediumllama-server -m /path/to/verified-model.gguf --port 8080
LM StudioEasySearch “meta-llama/Llama-3.2-3B-Instruct” → Load a Q4/INT4 build → Start Server
MLX LMMediummlx_lm.chat --model meta-llama/Llama-3.2-3B-Instruct
Watch out

Long advertised context still increases KV-cache memory; use modest context on small devices.