Microsoft

Phi 4 Reasoning 14B

Microsoft·14B·Phi·32K ctx·2026-08

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

Model size→Hardware requirement→Actual fit

Phi 4 Reasoning 14B needs ~12.2–22.2 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

12.2
8
Q4_K_M · Est. ~3 tok/sSlow · CPU offload

12 GB

RTX 4070 · 3060 12G

12.2
12
Q4_K_M · Est. ~6 tok/sSlow · CPU offload

16 GB

RTX 4080 · M-series 16G

14.2
16
Q5_K_M · Est. ~47 tok/sComfortable

24 GB

RTX 4090 · 3090

22.2
24
Q8_0 · Est. ~44 tok/sComfortable

48 GB

A6000 · 2×24G

22.2
48
Q8_0 · Est. ~44 tok/sComfortable

02  — Quantizations

QuantFile sizeEst. VRAM
Q4_K_M11 GB12.2 GB
Q5_K_M13 GB14.2 GB
Q8_021 GB22.2 GB

03  — Runtimes & use

llama.cpp · Ollama · LM Studio

General · Reasoning

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.