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.

1차 출처

총 파라미터12B

추정 메모리~9.4 GBINT4

컨텍스트128KText + Image

LicenseGemma140+ languages

01 / 추천 용도

vision
knowledge work
multilingual

02 / 추정 적합도

Mac Studio M5 Ultra · 96 GB여유 있음 Mac mini M6 · 24 GB여유 있음 MacBook Pro M5 Max · 64 GB여유 있음 Framework Desktop · Ryzen AI Max+ 395 · 128 GB여유 있음 Radeon AI PRO R9700 · 32 GB여유 있음

03 / 배포 환경

These are runtime examples, not validated installation recipes. Fit estimates assume quantized files; official repositories may hold full-precision weights. MLX, GGUF and TensorRT require matching conversions or engines.

OllamaEasyollama run gemma3:12b

llama.cppMediumllama-server -m /path/to/verified-model.gguf --port 9931

LM StudioEasySearch “google/gemma-3-12b-it” → Load a Q4/INT4 build → Start Server

MLX LMMediummlx_lm.server --model google/gemma-3-12b-it

주의

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