[Models](/models) / [Phi 4 14B](/models?family=Phi%204%2014B#models-gallery) / Phi 4

![Microsoft logo](/model-logos/microsoft.svg)

# Run Phi 4 on Mac

Phi 4 runs 100% private on Mac inside Private LLM — no internet connection required, no data sent to any server.

14.7BMac

## Download the quantized weights

These are the exact GPTQ-Int4 weights Private LLM runs for Phi 4, published on our Hugging Face org. They're standard weights you can load in any app that supports the format, not just Private LLM.

-   [Mac — GPTQ-Int4 →](https://huggingface.co/numen-tech/phi-4-GPTQ-Int4-D)

## Specifications

Parameters

14.7B

Context window

16K tokens

License

MIT

Quantization

GPTQ-Int4

Family

Phi 4 14B

## What Phi 4 is good at

general

Phi 4 is a 14 billion parameter language model from Microsoft Research, trained on high quality text to handle reasoning and logic tasks. It is designed for use in environments with limited memory or computing power and where fast responses are needed. The model performs well on benchmarks for math, science, and code generation compared to similar sized models.

## Which of your devices can run it

### Mac

Mac (Apple Silicon, 192GB)MacBook Pro (M4 Max, 128GB)Mac Studio / Pro (Apple Silicon, 96GB)MacBook Pro (M4 Max, 64GB)MacBook Pro (M4 Max, 48GB)MacBook Pro (M4 Max, 36GB)Mac (Apple Silicon, 32GB)MacBook Air (M4, 24GB)MacBook Air (M-series, 16GB)

Browse every model that fits [iPhone](/models/best-for-iphone) or [Mac](/models/best-for-mac).

## How to run Phi 4 in Private LLM

1.  Download Private LLM from the App Store.
2.  Open the in-app model library and choose Phi 4.
3.  Download the model once, then chat fully offline.

[![Download Private LLM on the App Store](/app-store/download-badge/en/download.svg)![Download Private LLM on the App Store](/app-store/download-badge/en/download-dark.svg)](/download)[Join our Discord](/discord)

## Variants & related models

![Meta logo](/model-logos/meta.svg)

### Airoboros l2 7b 3.0

6.7B

general

4K contextiPhone & iPad · Mac

[View details →](/models/airoboros-l2-7b-3.0)

![Mistral logo](/model-logos/mistral.svg)

### Airoboros M 7B

7.2B

general

33K contextMac

[View details →](/models/airoboros-m-7b-3.1.2)

![Meta logo](/model-logos/meta.svg)

### Cat Llama 3 70B Instruct

70.6B

general

8K contextMac

[View details →](/models/cat-llama-3-70b-instruct)

![Google Gemma logo](/model-logos/gemma.svg)

### FuseChat Gemma 2 9B Instruct

9.2B

general

8K contextiPhone & iPad · Mac

[View details →](/models/fusechat-gemma-2-9b-instruct)

![Meta logo](/model-logos/meta.svg)

### FuseChat Llama 3.1 8B Instruct

8B

general

131K contextiPhone & iPad · Mac

[View details →](/models/fusechat-llama-3.1-8b-instruct)

![Meta logo](/model-logos/meta.svg)

### FuseChat Llama 3.2 1B Instruct

1.2B

general

131K contextiPhone & iPad · Mac

[View details →](/models/fusechat-llama-3.2-1b-instruct)

## Frequently asked questions

-   Can I run Phi 4 on iPhone?
    
    Phi 4 is too large for current iPhones. It runs on Mac with enough unified memory inside Private LLM, fully offline.
    
-   Can I run Phi 4 on Mac?
    
    Yes. Phi 4 runs on Macs with enough unified memory, such as Mac (Apple Silicon, 192GB), MacBook Pro (M4 Max, 128GB), Mac Studio / Pro (Apple Silicon, 96GB), MacBook Pro (M4 Max, 64GB), MacBook Pro (M4 Max, 48GB), MacBook Pro (M4 Max, 36GB), Mac (Apple Silicon, 32GB), MacBook Air (M4, 24GB), MacBook Air (M-series, 16GB), fully on-device in Private LLM.
    
-   Does Phi 4 work offline?
    
    Yes. Once downloaded in Private LLM, Phi 4 runs 100% on-device — no internet connection, and nothing is sent to any server.
    
-   Is Phi 4 free to use?
    
    Private LLM is a one-time purchase with no subscription and no per-message cost. The models themselves are open source — once downloaded, they run offline with nothing to pay per use.
    

## Why run Phi 4 in Private LLM

Private LLM has run local AI on iPhone, iPad, and Mac since 2023, before Apple Intelligence existed. Inference happens on your device, so your Phi 4 conversations never reach a server. The part most apps gloss over is quantization, and that is exactly where on-device quality is won or lost. Most llama.cpp and MLX wrappers ship the same off-the-shelf 4-bit RTN weights. Private LLM ships GPTQ and OmniQuant quantization, tuned per model, and our 3-bit OmniQuant models match or beat those 4-bit RTN builds on the same Apple Silicon. Run the same model both ways and you feel it in the first reply. [See how our quantization works](/en/faq#Why-can-t-Private-LLM-load-models-directly-from-Hugging-Face).

### More from Numen

[![Slop or Not app icon](/app-icons/slop-or-not.png)

#### Slop or Not

Catch AI-written text on iPhone and Mac. The check runs on-device, so nothing you paste gets uploaded.

](https://slopornot.ai)[![Clean Links app icon](/app-icons/clean-links.png)

#### Clean Links

Strip trackers and clutter from any link before you share it. The cleanup happens on your device.

](https://cleanlinks.app)

Specifications and summary come from Phi 4's [Hugging Face model card](https://huggingface.co/microsoft/phi-4), released under the MIT license. Private LLM ships its own quantized models, built with GPTQ-Int4 quantization tuned per model, and isn't affiliated with the model's authors.