Qwen on Apple M3

Can I run Qwen3 235B-A22B on an Mac Studio M3 Ultra 256GB?

Yes. Qwen3 235B-A22B at Q4_K_M uses 134.8 GB of the 192.0 GB available on a Mac Studio M3 Ultra 256GB, and runs at about 24.7 tokens per second. You can push the context to 128K.

Runs comfortably

134.8 GB of 192.0 GB · 70%
0192 GB
Weights 132.8 GB
KV cache 1.5 GB
Runtime overhead 0.5 GB

Qwen3 235B-A22B at Q4_K_M leaves 57.2 GB spare on a Mac Studio M3 Ultra 256GB. There is room to raise the context length or move up a quantisation level.

Generation24.7tok/s
Prompt processing541tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Qwen3 235B-A22B 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 16.00 437.9 GB 439.9 GB Won't fit 1.42
INT8 / W8A8 8.50 232.3 GB 234.3 GB Won't fit 5.87
Q8_0 (GGUF) 8.50 232.3 GB 234.3 GB Won't fit 5.87
FP8 (E4M3) 8.00 218.9 GB 220.9 GB Won't fit 7.50
Q6_K 6.56 179.5 GB 181.5 GB Fits, but tight 65K 19.5
Q5_K_M 5.67 155.2 GB 157.2 GB Runs comfortably 128K 21.9
Q5_K_S 5.52 151.1 GB 153.1 GB Runs comfortably 128K 22.3
Q4_K_M 4.85 132.8 GB 134.8 GB Runs comfortably 128K 24.7
Q4_K_S 4.58 125.5 GB 127.5 GB Runs comfortably 128K 25.8
Q4_0 4.55 124.7 GB 126.7 GB Runs comfortably 128K 25.9
AWQ 4-bit 4.25 118.0 GB 120.0 GB Runs comfortably 128K 27.0
GPTQ 4-bit 4.25 118.0 GB 120.0 GB Runs comfortably 128K 27.0
MXFP4 4.25 118.0 GB 120.0 GB Runs comfortably 128K 27.0
IQ4_XS 4.25 116.5 GB 118.5 GB Runs comfortably 128K 27.2
Q3_K_M 3.91 107.2 GB 109.2 GB Runs comfortably 128K 28.9
IQ3_M 3.70 101.5 GB 103.5 GB Runs comfortably 128K 30.0
IQ3_XXS 3.06 84.1 GB 86.1 GB Runs comfortably 128K 34.2
Q2_K 2.63 72.4 GB 74.4 GB Runs comfortably 128K 37.7
IQ2_XXS 2.06 56.9 GB 58.9 GB Runs comfortably 128K 43.6
IQ1_M 1.75 48.4 GB 50.4 GB Runs comfortably 128K 47.6

Also worth checking