Run Phi-4 Mini 3.8B locally on iPhone, iPad & Mac
Microsoft model that punches above its size. Recent phones.
In privateSLM, Phi-4 Mini 3.8B is an everyday assistant for chat, writing, summarising and quick questions. You download the model once, and from then on every conversation runs entirely on your device — no cloud, no account, no tracking, and it keeps working with the radio off.
At a glance
| Model | Phi-4 Mini 3.8B |
|---|---|
| Category | Generalists |
| Download size | 2.49 GB (one time) |
| Quantization | Q4_K_M (GGUF) |
| Minimum device RAM | 6 GB |
| Runs offline | Yes — fully on-device after download |
| Source | Hugging Face |
Will it run on my device?
Phi-4 Mini 3.8B needs a device with at least 6 GB of RAM — that means iPhone 15 Pro and later, M-series iPads, and any Apple Silicon Mac. The privateSLM catalog states the RAM requirement for every model up front, so you know before you download.
How to run Phi-4 Mini 3.8B in privateSLM
- Get privateSLM on the App Store — one-time purchase, no subscription.
- Open Models, pick Phi-4 Mini 3.8B, and tap Download (2.49 GB, once).
- Select it as your active model and chat — airplane mode included.
On iPhones, iPads and Macs with Apple Intelligence, privateSLM can also answer instantly with the built-in system model — zero downloads. A catalog model like Phi-4 Mini 3.8B is for when you want this particular specialist, offline, under your control.
Related models
- LFM2.5 1.2B Instruct — 0.73 GB, needs 2 GB RAM
- Qwen2.5 1.5B Instruct — 0.99 GB, needs 3 GB RAM
- Gemma 2 2B Instruct — 1.71 GB, needs 4 GB RAM
- Llama 3.2 3B Instruct — 2.02 GB, needs 4 GB RAM
- Browse the full catalog →
FAQ
Does Phi-4 Mini 3.8B run offline on iPhone?
Yes. After a one-time 2.49 GB download inside privateSLM, Phi-4 Mini 3.8B runs entirely on-device via llama.cpp. Airplane mode works — nothing is ever sent to a server.
How much RAM does Phi-4 Mini 3.8B need?
A device with at least 6 GB of RAM — in practice iPhone 15 Pro and later, M-series iPads, and any Apple Silicon Mac.
Is Phi-4 Mini 3.8B free to use in privateSLM?
The model itself is open-weights (Q4_K_M GGUF) and the download is free. privateSLM is a one-time purchase — no subscription, no per-message fees.