OpenAI on Apple M4

Can I run gpt-oss 120B-A5.1B on an Mac Mini M4 Pro 48GB?

Not at Q4_K_M — it needs 67.1 GB against 36.0 GB available. You would need 2 of these cards.

Won't fit

67.1 GB of 36.0 GB · 186%
070 GB
Weights 66.0 GB
KV cache 0.6 GB
Runtime overhead 0.5 GB
Over the limit 31.1 GB

Short by 31.1 GB. You can run it with 19 of 36 layers on the Mac Mini M4 Pro 48GB and the rest in system RAM, at roughly 20.4 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation20.4tok/s
Prompt processing700tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of gpt-oss 120B-A5.1B on a Mac Mini M4 Pro 48GB

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

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 217.6 GB 218.6 GB Won't fit — 4.64
INT8 / W8A8 8.50 115.3 GB 116.4 GB Won't fit — 9.54
FP8 (E4M3) 8.00 108.8 GB 109.8 GB Won't fit — 10.3
Q6_K 6.56 89.2 GB 90.3 GB Won't fit — 13.4
Q5_K_M 5.67 77.1 GB 78.2 GB Won't fit — 16.2
Q5_K_S 5.52 75.1 GB 76.1 GB Won't fit — 16.6
Q4_K_M 4.85 66.0 GB 67.1 GB Won't fit — 20.4
Q4_K_S 4.58 62.4 GB 63.5 GB Won't fit — 22.2
Q4_0 4.55 62.0 GB 63.1 GB Won't fit — 22.3
AWQ 4-bit 4.25 59.4 GB 60.4 GB Won't fit — 23.9
GPTQ 4-bit 4.25 59.4 GB 60.4 GB Won't fit — 23.9
MXFP4 4.25 59.4 GB 60.4 GB Won't fit — 23.9
IQ4_XS 4.25 58.0 GB 59.0 GB Won't fit — 24.4
Q3_K_M 3.91 53.4 GB 54.4 GB Won't fit — 28.1
IQ3_M 3.70 50.6 GB 51.6 GB Won't fit — 30.6
IQ3_XXS 3.06 41.9 GB 43.0 GB Won't fit — 43.9
Q2_K 2.63 36.1 GB 37.2 GB Won't fit — 63.8
IQ2_XXS 2.06 28.5 GB 29.5 GB Runs comfortably 100K 86.0
IQ1_M 1.75 24.3 GB 25.4 GB Runs comfortably 128K 96.0

Also worth checking

gpt-oss 120B-A5.1B on every GPU
Full hardware tableEverything that runs on a Mac Mini M4 Pro 48GB
Full model tableWhy the KV cache is the part that bites
Guide