Rent a GPU

Can the Apple M1 (16GB) run Phi-3.5 Mini 3.8B?

Apple M1 (16GB) × Phi-3.5 Mini 3.8B8K context
YesRuns fully on the GPU
Memory at Q8_07.9 GB of 10.7 GB
Weights 4.1 GBKV cache 3.2 GBRuntime 0.6 GB
QuantQ8_0Near-lossless
Memory7.9 GBof 10.7 GB usable
Speed~10tokens per second
  1. FP16
  2. Q8_0
  3. Q6_K
  4. Q5_K_M
  5. Q4_K_M
  6. Q3_K_M
  7. Q2_K

Yes. The Apple M1 (16GB) can run Phi-3.5 Mini 3.8B fully in unified memory at Q8_0 (near-lossless), using about 7.9 GB of the 10.7 GB available, at roughly 10 tokens/second.

Try other context lengths

Run it with Ollama

With Ollama installed, one command downloads and starts Phi-3.5 Mini 3.8B:

ollama run phi3.5:3.8b

Ollama's default download is Q4_K_M (about 6.1 GB here). You have room for Q8_0, which is near-lossless: pick that GGUF in LM Studio or from Hugging Face.

Run it in the cloud

No Apple M1 (16GB)? Run Phi-3.5 Mini 3.8B in the cloud.

Rent a cloud GPU by the hour. No upfront cost, stop any time.

Rent a GPU on RunPodRent a GPU on Vast.ai

Memory needed by quantization

At 8,192 tokens of context. The black line marks the 10.7 GB the Apple M1 (16GB) can use. macOS lets the GPU use about 10.7 GB of the 16 GB unified memory by default.

QuantQualityMemoryFits?Speed
FP16Full quality11.5 GBToo big—
Q8_0Near-lossless7.9 GBFits~10 tok/s
Q6_KExcellent7 GBFits~13 tok/s
Q5_K_MVery good6.5 GBFits~15 tok/s
Q4_K_MGood, most popular6.1 GBFits~18 tok/s
Q3_K_MNoticeably degraded5.7 GBFits~22 tok/s
Q2_KHeavily degraded5.3 GBFits~28 tok/s

GPUs that run Phi-3.5 Mini 3.8B comfortably

Models that run well on the Apple M1 (16GB)

Frequently asked questions

Can Apple M1 (16GB) run Phi-3.5 Mini 3.8B?

Yes. The Apple M1 (16GB) can run Phi-3.5 Mini 3.8B fully in unified memory at Q8_0 (near-lossless), using about 7.9 GB of the 10.7 GB available, at roughly 10 tokens/second.

How much VRAM does Phi-3.5 Mini 3.8B need?

About 6.1 GB at Q4_K_M with 8,192 tokens of context, and about 11.5 GB at full FP16 precision.

How fast is Phi-3.5 Mini 3.8B on the Apple M1 (16GB)?

Roughly 10 tokens per second at Q8_0, estimated from the Apple M1 (16GB)'s 68 GB/s memory bandwidth.