Microsoft · Strong at math and step-by-step answers
Phi-4 Mini 3.8B on iPhone
- Balanced
- Tools (beta)
Phi-4 Mini is Microsoft’s 3.8-billion-parameter model, and in Priobs it runs fully on your iPhone. It punches above its size at maths, logic and coding help, and it shows its steps.
Free download · iPhone, iOS 26.0 or later
Specifications
- Company
- Microsoft
- Parameters
- 3.8B
- Download size
- 2.2 GB
- Minimum RAM
- 2.8 GB
- Category
- Balanced
- Tools
- Tools (beta)
- License
- MIT
- Format
- 4-bit quantized, MLX
Numbers come straight from the model catalogue inside the app. Phi-4 Mini 3.8B and its licence belong to Microsoft.
About Phi-4 Mini 3.8B
What is Phi-4 Mini?
Phi-4 Mini is part of Microsoft’s Phi family of small language models. Instead of sheer size, Phi relies on carefully chosen training data, which is why it does so well at structured problems. In Priobs it runs as a 4-bit quantized MLX build.
Who made it
Microsoft built the Phi series to show that small, well-trained models can hold their own on reasoning, coding and instruction following. Phi-4 Mini ships under the MIT licence, one of the most permissive open licences there is.
What it’s good at on iPhone
Phi-4 Mini is strong at maths and step-by-step answers:
- Maths problems and calculations with the working shown
- Logic puzzles and structured questions
- Coding help and explaining technical ideas
- Breaking a task into clear steps
Tool support for Phi-4 Mini 3.8B is in beta: it can use Priobs tools like calendar and reminders, but expect the odd miss.
It is not the most capable model in the library. That is Qwen 3 4B, which is also better with tools. Pick Phi-4 Mini when your question is mostly numbers, logic or code.
Which iPhones can run it
Phi-4 Mini 3.8B needs at least 2.8 GB of memory. Priobs checks your iPhone, hides models it does not have the memory for, and warns you if a model is close to the limit and may be unstable. As a rough guide, an iPhone 12 or later with 4 GB of RAM can run it, and an iPhone 14 or newer (6 GB or more) gives it more room.
Setup
How to get it in Priobs.
One download over Wi-Fi and you are offline-ready for good.
Open Priobs and go to the Model Library.
Find Phi-4 Mini and tap Download (2.2 GB, Wi-Fi recommended).
Select it as your active model. It is ready to use offline.
FAQ
Phi-4 Mini 3.8B, answered.
Is Phi-4 Mini the best model in Priobs?
Does Phi-4 Mini support tools?
Will Phi-4 Mini drain my battery faster?
Does Phi-4 Mini send my prompts to Microsoft?
Can I switch to a smaller model if Phi-4 Mini feels slow?
Other models
You can keep several installed at once.
Switching models takes one tap. A common setup: a small, fast model for quick questions and a bigger one for anything that needs more depth.
Qwen 2.5 0.5B
Alibaba
FastestFastest responses, good for simple tasks
288 MB · 512 MB RAM
LFM2 1.2B
Liquid AI
FastestToolsBuilt for speed on phones, snappy short replies
663 MB · 862 MB RAM
Gemma 3 1B
Google
BalancedBalanced speed and quality
767 MB · 1 GB RAM
Qwen 2.5 1.5B
Alibaba
BalancedStrong multilingual support
878 MB · 1.2 GB RAM
Qwen 3 1.7B
Alibaba
BalancedToolsReasoningCompact model that thinks before answering
982 MB · 1.2 GB RAM
Llama 3.2 3B
Meta
BalancedToolsGreat for writing and everyday questions
1.8 GB · 2.3 GB RAM
Qwen 3 4B
Alibaba
Most CapableToolsReasoningMost capable, best for complex tasks
2.3 GB · 2.9 GB RAM
Qwen 3 4B Thinking
Alibaba
Most CapableTools (beta)ReasoningWorks through problems step by step, slower to answer
2.3 GB · 2.9 GB RAM
Gemma 4 E2B
Google
BalancedTools (beta)Google’s latest with strong reasoning
3.6 GB · 4.5 GB RAM
Apple Intelligence
Apple
Built inToolsBuilt into your iPhone, nothing to download
No download, built into iOS