Gemma on Apple M3

Can I run Gemma 3 12B on an Mac Studio M3 Ultra 256GB?

Yes. Gemma 3 12B at Q4_K_M uses 8.3 GB of the 192.0 GB available on a Mac Studio M3 Ultra 256GB, and runs at about 61.8 tokens per second. You can push the context to 128K.

Runs comfortably

8.3 GB of 192.0 GB · 4%
0192 GB
Weights 7.0 GB
KV cache 0.8 GB
Runtime overhead 0.5 GB

Gemma 3 12B at Q4_K_M leaves 183.7 GB spare on a Mac Studio M3 Ultra 256GB. There is room to raise the context length or move up a quantisation level.

Generation61.8tok/s
Prompt processing981tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Gemma 3 12B on a Mac Studio M3 Ultra 256GB

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 22.7 GB 24.1 GB Runs comfortably 128K 20.2
INT8 / W8A8 8.50 11.8 GB 13.2 GB Runs comfortably 128K 37.7
Q8_0 (GGUF) 8.50 11.8 GB 13.2 GB Runs comfortably 128K 37.7
FP8 (E4M3) 8.00 11.4 GB 12.7 GB Runs comfortably 128K 39.2
Q6_K 6.56 9.3 GB 10.6 GB Runs comfortably 128K 47.2
Q5_K_M 5.67 8.1 GB 9.4 GB Runs comfortably 128K 54.1
Q5_K_S 5.52 7.8 GB 9.2 GB Runs comfortably 128K 55.4
AWQ 4-bit 4.25 7.4 GB 8.7 GB Runs comfortably 128K 58.3
GPTQ 4-bit 4.25 7.4 GB 8.7 GB Runs comfortably 128K 58.3
MXFP4 4.25 7.4 GB 8.7 GB Runs comfortably 128K 58.3
Q4_K_M 4.85 7.0 GB 8.3 GB Runs comfortably 128K 61.8
Q4_K_S 4.58 6.6 GB 7.9 GB Runs comfortably 128K 64.7
Q4_0 4.55 6.6 GB 7.9 GB Runs comfortably 128K 65.1
IQ4_XS 4.25 6.2 GB 7.5 GB Runs comfortably 128K 68.8
Q3_K_M 3.91 5.7 GB 7.1 GB Runs comfortably 128K 73.5
IQ3_M 3.70 5.5 GB 6.8 GB Runs comfortably 128K 76.8
IQ3_XXS 3.06 4.6 GB 6.0 GB Runs comfortably 128K 88.8
Q2_K 2.63 4.1 GB 5.4 GB Runs comfortably 128K 99.2
IQ2_XXS 2.06 3.3 GB 4.7 GB Runs comfortably 128K 117
IQ1_M 1.75 2.9 GB 4.3 GB Runs comfortably 128K 131

Also worth checking