Phi-4 14B
Phi — 14B params — Q4_K_M, local via Ollama. Dataset updated 2026-09-03.
PARAMETERS
14B
FORMAT
Q4_K_M
MIN MEMORY
24 GB
BEST FOR
Coding, Quality
Memory math
Weights (Q4_K_M)
11.5 GB
+ KV cache (16k)
~3.0 GB
Total at 16k
~14.5 GB
Comfortable from
24 GB
KV = fp16 estimate (q8_0 cache roughly halves it). "Comfortable" = weights + KV within the engine's tiered budget (~70-85% of memory).
Hardware snapshot
Go deeper
Every Phi-4 14B quant by real GGUF file sizeAlso tracked: Phi-4 14B (Q8) (Q8_0, 14.5 GB, min 24 GB)
More Phi models
Frequently asked questions
How much memory does Phi-4 14B need?
11.5 GB for the Q4_K_M weights, plus ~3.0 GB of KV-cache at 16k context — about 14.5 GB total. Comfortable from 24 GB of VRAM or unified memory.
Does Phi-4 14B run on a Mac?
Yes — from the MacBook Air M5 24GB (~14 tok/s est.). Unified memory means the RAM budget is the only limit.
What is the cheapest GPU for Phi-4 14B?
The AMD Radeon RX 7900 XT is the cheapest tracked card that runs Phi-4 14B comfortably — ~46 tok/s est. at ~$550 used.
Cite this page
ModelFit: Phi-4 14B — specs, memory math and hardware verdicts. https://modelfit.io/models/phi4-14b/ (dataset updated 2026-09-03, CC BY 4.0).