Gemma on Apple M4

Can I run Gemma 3 4B on an Mac Studio M4 Max 128GB?

Yes. Gemma 3 4B at Q4_K_M uses 3.3 GB of the 96.0 GB available on a Mac Studio M4 Max 128GB, and runs at about 128 tokens per second. You can push the context to 128K.

Runs comfortably

3.3 GB of 96.0 GB · 3%
096 GB
Weights 2.5 GB
KV cache 0.3 GB
Runtime overhead 0.5 GB

Gemma 3 4B at Q4_K_M leaves 92.7 GB spare on a Mac Studio M4 Max 128GB. There is room to raise the context length or move up a quantisation level.

Generation128tok/s
Prompt processing1660tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Gemma 3 4B on a Mac Studio M4 Max 128GB

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 8.0 GB 8.8 GB Runs comfortably 128K 42.4
INT8 / W8A8 8.50 4.1 GB 4.9 GB Runs comfortably 128K 80.2
Q8_0 (GGUF) 8.50 4.1 GB 4.9 GB Runs comfortably 128K 80.2
FP8 (E4M3) 8.00 4.0 GB 4.8 GB Runs comfortably 128K 82.1
Q6_K 6.56 3.3 GB 4.1 GB Runs comfortably 128K 98.7
AWQ 4-bit 4.25 3.0 GB 3.8 GB Runs comfortably 128K 106
GPTQ 4-bit 4.25 3.0 GB 3.8 GB Runs comfortably 128K 106
MXFP4 4.25 3.0 GB 3.8 GB Runs comfortably 128K 106
Q5_K_M 5.67 2.8 GB 3.6 GB Runs comfortably 128K 113
Q5_K_S 5.52 2.8 GB 3.6 GB Runs comfortably 128K 116
Q4_K_M 4.85 2.5 GB 3.3 GB Runs comfortably 128K 128
Q4_K_S 4.58 2.4 GB 3.2 GB Runs comfortably 128K 133
Q4_0 4.55 2.4 GB 3.1 GB Runs comfortably 128K 134
IQ4_XS 4.25 2.2 GB 3.0 GB Runs comfortably 128K 141
Q3_K_M 3.91 2.1 GB 2.9 GB Runs comfortably 128K 149
IQ3_M 3.70 2.0 GB 2.8 GB Runs comfortably 128K 155
IQ3_XXS 3.06 1.7 GB 2.5 GB Runs comfortably 128K 176
Q2_K 2.63 1.5 GB 2.3 GB Runs comfortably 128K 194
IQ2_XXS 2.06 1.3 GB 2.1 GB Runs comfortably 128K 223
IQ1_M 1.75 1.2 GB 2.0 GB Runs comfortably 128K 243

Also worth checking

Gemma 3 4B on every GPU
Full hardware tableEverything that runs on a Mac Studio M4 Max 128GB
Full model tableWhy the KV cache is the part that bites
Guide