Cohere on Apple M4

Can I run Command R 35B on an Mac Mini M4 Pro 48GB?

Yes. Command R 35B at Q4_K_M uses 30.7 GB of the 36.0 GB available on a Mac Mini M4 Pro 48GB, and runs at about 7.00 tokens per second. You can push the context to 12K.

Runs comfortably

30.7 GB of 36.0 GB · 85%
036 GB
Weights 20.1 GB
KV cache 10.0 GB
Runtime overhead 0.6 GB

Command R 35B at Q4_K_M leaves 5.3 GB spare on a Mac Mini M4 Pro 48GB. There is room to raise the context length or move up a quantisation level.

Generation7.00tok/s
Prompt processing102tok/s
Max context12Ktokens
KV per 1K tokens0GB

Every quantisation of Command R 35B 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 65.2 GB 75.7 GB Won't fit — 0.82
INT8 / W8A8 8.50 33.7 GB 44.3 GB Won't fit 1K 2.50
Q8_0 (GGUF) 8.50 33.7 GB 44.3 GB Won't fit 1K 2.50
FP8 (E4M3) 8.00 32.6 GB 43.2 GB Won't fit 2K 2.69
Q6_K 6.56 26.7 GB 37.3 GB Won't fit 7K 4.75
Q5_K_M 5.67 23.1 GB 33.7 GB Fits, but tight 10K 6.24
AWQ 4-bit 4.25 23.0 GB 33.6 GB Fits, but tight 10K 6.25
GPTQ 4-bit 4.25 23.0 GB 33.6 GB Fits, but tight 10K 6.25
MXFP4 4.25 23.0 GB 33.6 GB Fits, but tight 10K 6.25
Q5_K_S 5.52 22.5 GB 33.1 GB Fits, but tight 10K 6.38
Q4_K_M 4.85 20.1 GB 30.7 GB Runs comfortably 12K 7.00
Q4_K_S 4.58 19.1 GB 29.7 GB Runs comfortably 13K 7.28
Q4_0 4.55 19.0 GB 29.6 GB Runs comfortably 13K 7.31
IQ4_XS 4.25 17.9 GB 28.5 GB Runs comfortably 14K 7.65
Q3_K_M 3.91 16.7 GB 27.3 GB Runs comfortably 15K 8.08
IQ3_M 3.70 15.9 GB 26.5 GB Runs comfortably 16K 8.37
IQ3_XXS 3.06 13.7 GB 24.2 GB Runs comfortably 17K 9.40
Q2_K 2.63 12.1 GB 22.7 GB Runs comfortably 19K 10.2
IQ2_XXS 2.06 10.1 GB 20.7 GB Runs comfortably 20K 11.6
IQ1_M 1.75 9.0 GB 19.5 GB Runs comfortably 21K 12.5

Also worth checking

Command R 35B 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