Gemma 3 27B

by Google · gemma-3 family

27B

parameters

text-generation code-generation reasoning multilingual vision math creative-writing summarization

Gemma 3 27B is the flagship of the Gemma 3 family — a multimodal powerhouse that competes with models in the 65B+ class on many benchmarks. It supports 128K context and handles text, images, and complex reasoning tasks with high quality. At Q4 it needs about 20 GB VRAM, fitting well on a RTX 3090/4090 or a Mac with 24 GB+ unified memory. An excellent choice for users who want near-frontier performance from a single consumer GPU.

Quick Start with Ollama

ollama run 27b-it-q4_K_M
Resources Ollama Hugging Face Official Page
Creator Google
Parameters 27B
Architecture transformer-decoder
Context 128K tokens
Released Mar 12, 2025
License Gemma Terms of Use
Ollama gemma3:27b

Quantization Options

Format File Size VRAM Required Quality Ollama Tag
Q4_K_M rec 17 GB 20 GB
27b-it-q4_K_M
Q8_0 30 GB 34 GB
27b-it-q8_0
F16 55 GB 60 GB
27b-it-fp16

Compatible Hardware

for Q4_K_M (20 GB VRAM)

Compatible Hardware

Hardware VRAM Type Fit Est. Speed
Mac Studio M4 Ultra 512GB 512 GB mac Runs ~41 tok/s
Mac Pro M2 Ultra 192GB 192 GB mac Runs ~40 tok/s
Mac Studio M4 Ultra 192GB 192 GB mac Runs ~41 tok/s
Mac Studio M4 Max 128GB 128 GB mac Runs ~27 tok/s
MacBook Pro M4 Max 128GB 128 GB mac Runs ~27 tok/s
MacBook Pro M3 Max 96GB 96 GB mac Runs ~20 tok/s
Mac mini M4 Pro 64GB 64 GB mac Runs ~14 tok/s
Mac Studio M4 Max 64GB 64 GB mac Runs ~27 tok/s
MacBook Pro M4 Max 64GB 64 GB mac Runs ~27 tok/s
Mac mini M4 Pro 48GB 48 GB mac Runs ~14 tok/s
MacBook Pro M3 Max 48GB 48 GB mac Runs ~20 tok/s
MacBook Pro M4 Max 48GB 48 GB mac Runs ~27 tok/s
MacBook Pro M4 Pro 48GB 48 GB mac Runs ~14 tok/s
Mac Studio M4 Max 36GB 36 GB mac Runs ~27 tok/s
MacBook Pro M3 Pro 36GB 36 GB mac Runs ~8 tok/s
NVIDIA GeForce RTX 5090 32 GB gpu Runs ~90 tok/s
iMac M4 32GB 32 GB mac Runs ~6 tok/s
Mac mini M4 32GB 32 GB mac Runs ~6 tok/s
MacBook Air M4 32GB 32 GB mac Runs ~6 tok/s
AMD Radeon RX 7900 XTX 24 GB gpu Runs ~48 tok/s
NVIDIA GeForce RTX 3090 24 GB gpu Runs ~47 tok/s
NVIDIA GeForce RTX 4090 24 GB gpu Runs ~50 tok/s
iMac M3 24GB 24 GB mac Runs ~5 tok/s
Mac mini M2 24GB 24 GB mac Runs ~5 tok/s
Mac mini M4 Pro 24GB 24 GB mac Runs ~14 tok/s
MacBook Air M2 24GB 24 GB mac Runs ~5 tok/s
MacBook Air M4 24GB 24 GB mac Runs ~6 tok/s
MacBook Pro M4 Pro 24GB 24 GB mac Runs ~14 tok/s
AMD Radeon RX 7900 XT 20 GB gpu CPU Offload ~40 tok/s
MacBook Pro M3 Pro 18GB 18 GB mac CPU Offload ~8 tok/s
AMD Radeon RX 6800 XT 16 GB gpu CPU Offload ~26 tok/s
AMD Radeon RX 7800 XT 16 GB gpu CPU Offload ~31 tok/s
Intel Arc A770 16 GB gpu CPU Offload ~28 tok/s
NVIDIA GeForce RTX 4060 Ti 16GB 16 GB gpu CPU Offload ~14 tok/s
NVIDIA GeForce RTX 4070 Ti Super 16 GB gpu CPU Offload ~34 tok/s
NVIDIA GeForce RTX 4080 Super 16 GB gpu CPU Offload ~37 tok/s
NVIDIA GeForce RTX 4080 16 GB gpu CPU Offload ~36 tok/s
NVIDIA GeForce RTX 5070 Ti 16 GB gpu CPU Offload ~45 tok/s
NVIDIA GeForce RTX 5080 16 GB gpu CPU Offload ~48 tok/s
iMac M1 16GB 16 GB mac CPU Offload ~3 tok/s
iMac M4 16GB 16 GB mac CPU Offload ~6 tok/s
Mac mini M1 16GB 16 GB mac CPU Offload ~3 tok/s
Mac mini M4 16GB 16 GB mac CPU Offload ~6 tok/s
MacBook Air M2 16GB 16 GB mac CPU Offload ~5 tok/s
MacBook Air M3 16GB 16 GB mac CPU Offload ~5 tok/s
MacBook Air M4 16GB 16 GB mac CPU Offload ~6 tok/s
MacBook Pro M1 16GB 16 GB mac CPU Offload ~3 tok/s
MacBook Pro M2 Pro 16GB 16 GB mac CPU Offload ~10 tok/s
17 hardware device(s) cannot run this model configuration.

Benchmark Scores

78.5
mmlu