Microsoft

Phi 4 Mini Reasoning 3.8B

Microsoft·3.8B·Phi·128K ctx·2026-08

Check on my hardwareAdd to compare →Ollama →Hugging Face →

Model size→Hardware requirement→Actual fit

Phi 4 Mini Reasoning 3.8B needs ~4.4–7.3 GB of VRAM depending on quantization. The bar is what it needs; the marker is the card's VRAM. Speed figures below are planning estimates.

8 GB

RTX 4060 · 3060 Ti

5
8
Q5_K_M · Est. ~20 tok/sComfortable

12 GB

RTX 4070 · 3060 12G

7.3
12
Q8_0 · Est. ~24 tok/sComfortable

16 GB

RTX 4080 · M-series 16G

7.3
16
Q8_0 · Est. ~37 tok/sComfortable

24 GB

RTX 4090 · 3090

7.3
24
Q8_0 · Est. ~57 tok/sComfortable

48 GB

A6000 · 2×24G

7.3
48
Q8_0 · Est. ~57 tok/sComfortable

02  — Quantizations

QuantFile sizeEst. VRAM
Q4_K_M3.2 GB4.4 GB
Q5_K_M3.8 GB5 GB
Q8_06.1 GB7.3 GB

03  — Runtimes & use

llama.cpp · Ollama · LM Studio

General · Reasoning · Fast · Privacy

Memory and speed figures are planning estimates, not measured benchmarks. Memory includes a fixed allowance for KV cache and runtime overhead. Long-context prompts push VRAM use above the numbers shown here.