Text model · Phi-4

Phi-4-mini-reasoning: RAM and VRAM requirements

Phi-4-mini-reasoning needs about 3.6 GB to run at Q4_K_M (the Q4_K_M GGUF file is ~2.32 GB to download; KV cache and overhead add the rest), or about 5.1 GB at Q8_0. The lightest hardware that runs it is Apple M1 (8GB).

Phi-4 family · 3.8B params · released Apr 2025.

Shopping for hardware? See what runs Phi-4-mini-reasoning →

License MIT · Commercial OK ↓ 84.6K/mo ♥ 240 on HuggingFace

Q4_K_M GGUF

2.32 GB

Q8_0 GGUF

3.8 GB

Memory @ Q4 (4k)

~3.6 GB

Context

128 k

Q4_K_M and Q8_0 are GGUF download (file) sizes. Memory @ Q4 includes the KV cache and overhead, so it runs larger than the file on disk.

Will it run on your device?

Phi-4-mini-reasoning runs on 40 of 40 tracked devices at Q4_K_M.

35 run well 5 tight fit 0 too small

Fit check Q4_K_M

Your device

Apple M4 (16GB)

Yes, it runs fast

needs 3.6 GB usable 10.5 GB

See the full breakdown

$ollama run phi4-mini-reasoning:latest

Memory breakdown

Weights (Q4_K_M)2.32 GB

KV cache (4k)0.5 GB

Overhead0.8 GB

Total3.6 GB

How context length changes it

4k context ~3.6 GB 32k context ~6.8 GB 128k context ~18.1 GB

Phi-4-mini-reasoning is small enough that the context window is the thing to size for: 128k needs ~18.1 GB versus ~3.6 GB at 4k, so pick the window you actually use.

Speed drops too: every token re-reads the KV cache, so at 128k context Phi-4-mini-reasoning generates at roughly ~13% of its short-context speed. How this is estimated.

Quantization sizes

GGUF quantson disk

Quantization	Size on disk
Q2_K	1.6 GB est
Q3_K_M	1.9 GB est
Q4_K_M (default)	2.32 GB
Q5_K_M	2.7 GB est
Q6_K	3.1 GB est
Q8_0	3.8 GB
FP16	7.68 GB

Lower quant = smaller and faster, slightly lower quality. Q4_K_M is the common default.

Run it

Ollama

$ ollama run phi4-mini-reasoning:latest

llama.cpp

$ llama-cli -hf unsloth/Phi-4-mini-reasoning-GGUF:Q4_K_M

LM Studio

$ lms get unsloth/Phi-4-mini-reasoning-GGUF

Which devices can run Phi-4-mini-reasoning?

Apple Silicon Macs

RAM-only laptops

iPhone & iPad

Android

NVIDIA GPUs

AMD GPUs

Same job, different size

Pick by what fits your memory: step down to free up VRAM, or step up if you have headroom.

Next size down Qwen2.5-VL 3B 3.75B Next size up Phi-3.5-mini 3.8B 3.82B

Similar models

Head-to-head

Phi-4-mini-reasoning vs Phi-4-reasoning

FAQ

How much VRAM or RAM does Phi-4-mini-reasoning need?

At Q4_K_M, Phi-4-mini-reasoning needs about 3.6 GB (weights ~2.32 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~5.1 GB.

What is the Q4_K_M GGUF file size of Phi-4-mini-reasoning?

The Q4_K_M GGUF file is about 2.32 GB to download, and the Q8_0 GGUF is about 3.8 GB. That is the weights file on disk; to run it you also need room for the KV cache and overhead, so budget ~3.6 GB of memory at Q4_K_M.

Can Phi-4-mini-reasoning run on a laptop?

Yes, Phi-4-mini-reasoning fits on a 16 GB machine at Q4_K_M and runs on Apple Silicon or a 12 GB+ GPU comfortably.

Can I use Phi-4-mini-reasoning commercially?

Yes. Phi-4-mini-reasoning is licensed MIT, which permits commercial use.

Understand the numbers

Short guides to the ideas behind Phi-4-mini-reasoning's memory and quant figures.

Microsoft's tiny reasoning model: 3.8B, MIT, runs on almost anything.

Sources

Last validated 2026-08-03. Memory figures are estimates. See methodology.