← Back to all models

Phi 3.5 MoE instruct

Chat model by Microsoft · 41.87B parameters

Chip
Unified Memory
Bandwidth 307 GB/s
Available for Models ~21 GB

Everything this chip runs, machine by machine: MacBook Pro 14" M5 Pro, MacBook Pro 16" M5 Pro

Tight Fit

Phi 3.5 MoE instruct can run but will be tight at 78% of RAM. Consider a lower quantization or closing other apps.

tok/s

Q2_K · 18.7 GB · Benchmark needed

Quantization Options

QuantizationMemorySpeedFits?
Q8_0 47.2 GB~23 tok/s
Q6_K 36.7 GB~30 tok/s
Q5_K_M 31.7 GB~35 tok/s
Q4_K_M 27.5 GB~41 tok/s
Q3_K_M 23.3 GB~50 tok/s
Q2_K Recommended 18.7 GB~65 tok/s

Specifications

Parameters41.9B
Architecturephimoe
Context Length131K
CategoryChat
Mixture of Experts 16 experts, 2 active
Formatgguf
Minimum RAM23.4 GB
Licensemit
HuggingFace Downloads134,995

Which Mac Can Run Phi 3.5 MoE instruct?

Minimum Mac

MacBook Pro 14" — M5 Max 48GB

Q6_K · benchmark needed · 77% RAM

MacQuanttok/sRAM %
Mac Studio · M3 Ultra 96GB Q8_0 49%
Mac Studio · M3 Ultra 256GB Q8_0 18%
Mac Studio · M3 Ultra 512GB Q8_0 9%
MacBook Pro 14" · M5 Max 48GB Q6_K 77%
MacBook Pro 16" · M5 Max 48GB Q6_K 77%
Mac Studio · M1 Ultra 64GB Q8_0 74%
Mac Studio · M1 Ultra 128GB Q8_0 37%
Mac Studio · M2 Ultra 64GB Q8_0 74%
Mac Studio · M2 Ultra 128GB Q8_0 37%
Mac Studio · M2 Ultra 192GB Q8_0 25%
MacBook Pro 14" · M4 Max 48GB Q6_K 77%
MacBook Pro 16" · M4 Max 48GB Q6_K 77%
Mac Studio · M4 Max 48GB Q6_K 77%
MacBook Pro 14" · M5 Max 64GB Q8_0 74%
MacBook Pro 14" · M5 Max 128GB Q8_0 37%
MacBook Pro 16" · M5 Max 64GB Q8_0 74%
MacBook Pro 16" · M5 Max 128GB Q8_0 37%
MacBook Pro 14" · M4 Max 64GB Q8_0 74%
MacBook Pro 14" · M4 Max 128GB Q8_0 37%
MacBook Pro 16" · M4 Max 64GB Q8_0 74%
MacBook Pro 16" · M4 Max 128GB Q8_0 37%
Mac Studio · M4 Max 64GB Q8_0 74%
Mac Studio · M4 Max 128GB Q8_0 37%
MacBook Pro 14" · M3 Max 48GB Q6_K 77%
MacBook Pro 16" · M3 Max 48GB Q6_K 77%
MacBook Pro 14" · M5 Pro 48GB Q6_K 77%
MacBook Pro 16" · M5 Pro 48GB Q6_K 77%
MacBook Pro 14" · M1 Max 64GB Q8_0 74%
MacBook Pro 14" · M2 Max 64GB Q8_0 74%
MacBook Pro 14" · M2 Max 96GB Q8_0 49%
MacBook Pro 14" · M3 Max 64GB Q8_0 74%
MacBook Pro 14" · M3 Max 96GB Q8_0 49%
MacBook Pro 14" · M3 Max 128GB Q8_0 37%
MacBook Pro 16" · M1 Max 64GB Q8_0 74%
MacBook Pro 16" · M2 Max 64GB Q8_0 74%
MacBook Pro 16" · M2 Max 96GB Q8_0 49%
MacBook Pro 16" · M3 Max 64GB Q8_0 74%
MacBook Pro 16" · M3 Max 96GB Q8_0 49%
MacBook Pro 16" · M3 Max 128GB Q8_0 37%
Mac Studio · M1 Max 64GB Q8_0 74%
Mac Studio · M2 Max 64GB Q8_0 74%
Mac Studio · M2 Max 96GB Q8_0 49%
MacBook Pro 14" · M4 Pro 48GB Q6_K 77%
MacBook Pro 16" · M4 Pro 48GB Q6_K 77%
Mac mini · M4 Pro 48GB Q6_K 77%
MacBook Pro 14" · M5 Pro 64GB Q8_0 74%
MacBook Pro 16" · M5 Pro 64GB Q8_0 74%
Mac mini · M4 Pro 64GB Q8_0 74%
MacBook Pro 14" · M5 Max 36GB Q4_K_M 76%
MacBook Pro 16" · M5 Max 36GB Q4_K_M 76%
MacBook Pro 14" · M4 Max 36GB Q4_K_M 76%
MacBook Pro 16" · M4 Max 36GB Q4_K_M 76%
Mac Studio · M4 Max 36GB Q4_K_M 76%
MacBook Pro 14" · M5 Pro 24GB Q2_K 78%
MacBook Pro 16" · M5 Pro 24GB Q2_K 78%
MacBook Pro 14" · M4 Pro 24GB Q2_K 78%
MacBook Pro 16" · M4 Pro 24GB Q2_K 78%
Mac mini · M4 Pro 24GB Q2_K 78%
MacBook Pro 14" · M1 Max 32GB Q4_K_M 86%
MacBook Pro 14" · M2 Max 32GB Q4_K_M 86%
MacBook Pro 14" · M3 Max 36GB Q4_K_M 76%
MacBook Pro 16" · M1 Max 32GB Q4_K_M 86%
MacBook Pro 16" · M2 Max 32GB Q4_K_M 86%
MacBook Pro 16" · M3 Max 36GB Q4_K_M 76%
Mac Studio · M1 Max 32GB Q4_K_M 86%
Mac Studio · M2 Max 32GB Q4_K_M 86%
MacBook Air · M5 24GB Q2_K 78%
MacBook Pro 14" · M5 24GB Q2_K 78%
MacBook Pro 14" · M1 Pro 32GB Q4_K_M 86%
MacBook Pro 14" · M2 Pro 32GB Q4_K_M 86%
MacBook Pro 16" · M1 Pro 32GB Q4_K_M 86%
MacBook Pro 16" · M2 Pro 32GB Q4_K_M 86%
Mac mini · M2 Pro 32GB Q4_K_M 86%
MacBook Air · M4 24GB Q2_K 78%
MacBook Pro 14" · M4 24GB Q2_K 78%
Mac mini · M4 24GB Q2_K 78%
iMac · M4 24GB Q2_K 78%
MacBook Air · M2 24GB Q2_K 78%
MacBook Air · M3 24GB Q2_K 78%
MacBook Pro 14" · M3 24GB Q2_K 78%
Mac mini · M2 24GB Q2_K 78%
iMac · M3 24GB Q2_K 78%
MacBook Air · M5 32GB Q4_K_M 86%
MacBook Pro 14" · M5 32GB Q4_K_M 86%
MacBook Pro 14" · M3 Pro 36GB Q4_K_M 76%
MacBook Pro 16" · M3 Pro 36GB Q4_K_M 76%
MacBook Air · M4 32GB Q4_K_M 86%
MacBook Pro 14" · M4 32GB Q4_K_M 86%
Mac mini · M4 32GB Q4_K_M 86%
iMac · M4 32GB Q4_K_M 86%
MacBook Pro 14" · M1 Pro 16GB 117%
MacBook Pro 14" · M2 Pro 16GB 117%
MacBook Pro 16" · M1 Pro 16GB 117%
MacBook Pro 16" · M2 Pro 16GB 117%
Mac mini · M2 Pro 16GB 117%
MacBook Air · M5 16GB 117%
MacBook Pro 14" · M5 16GB 117%
MacBook Pro 14" · M3 Pro 18GB 104%
MacBook Pro 16" · M3 Pro 18GB 104%
MacBook Air · M4 16GB 117%
MacBook Pro 14" · M4 16GB 117%
Mac mini · M4 16GB 117%
iMac · M4 16GB 117%
MacBook Air · M2 8GB 234%
MacBook Air · M2 16GB 117%
MacBook Air · M3 8GB 234%
MacBook Air · M3 16GB 117%
MacBook Pro 14" · M3 8GB 234%
MacBook Pro 14" · M3 16GB 117%
Mac mini · M2 8GB 234%
Mac mini · M2 16GB 117%
iMac · M3 8GB 234%
iMac · M3 16GB 117%
MacBook Air · M1 8GB 234%
MacBook Air · M1 16GB 117%
Mac mini · M1 8GB 234%
Mac mini · M1 16GB 117%
iMac · M1 8GB 234%
iMac · M1 16GB 117%

Run Phi 3.5 MoE instruct locally on your Mac

ToolPiper downloads, manages, and runs models with one click. Apple Silicon optimized.

Get ToolPiper — Free