LocalForge
Back to guide
MODEL / GOOGLE

Gemma 3 12B

Google reports an INT4 path on an 8 GB laptop GPU, but context and runtime overhead can make 8 GB tight.

Primary source
Total parameters12B
Estimated memory~9.4 GBINT4
Context128KText + Image
LicenseGemma140+ languages
01 / BEST FOR
  • vision
  • knowledge work
  • multilingual
02 / ESTIMATED FIT
16 GB CPU desktopTightApple silicon · 16 GBTightNVIDIA · 12 GBTightNVIDIA · 16 GBComfortableApple silicon · 32 GBComfortable
03 / DEPLOY WITH
OllamaEasyollama run gemma3:12b
llama.cppMediumllama-server -m /path/to/verified-model.gguf --port 8080
LM StudioEasySearch “google/gemma-3-12b-it” → Load a Q4/INT4 build → Start Server
MLX LMMediummlx_lm.chat --model google/gemma-3-12b-it
Watch out

Google reports an INT4 path on an 8 GB laptop GPU, but context and runtime overhead can make 8 GB tight.