Microsoft · Dense model · 2024-12

Phi-4 14B: how much VRAM?

In Q4_K_M with 8k tokens of context, Phi-4 14B needs about 11 GB of memory. The cheapest option: RTX 3060 12 GB (€1,180 for the complete machine).

Model sheet

Parameters15 B

Active / token15 B

Cache / 1k tok.205 MB

ollama run phi4

Weights on Hugging Face: microsoft/phi-4

Configure the cheapest server →Compare GPUs

VRAM by precision and context

Precision	4k	8k	32k	128k
FP16	32 GB	33 GB	38 GB	57 GB
FP8	17 GB	17 GB	22 GB	41 GB
Q8_0	18 GB	18 GB	23 GB	42 GB
Q6_K	14 GB	15 GB	19 GB	39 GB
Q4_K_M	10 GB	11 GB	16 GB	35 GB
Q3_K_M	9 GB	10 GB	14 GB	34 GB

One user. Each simultaneous user adds their own context cache.

Frequently asked questions

How much VRAM does Phi-4 14B need?

With an 8k-token context and one user, Phi-4 14B needs about 11 GB in Q4, 18 GB in Q8 and 33 GB in FP16.

What is the cheapest hardware to run Phi-4 14B?

In Q4, the cheapest option in our catalogue to run Phi-4 14B is RTX 3060 12 GB, about €1,180 for the complete machine.

How fast does Phi-4 14B run?

Expect around 22 tokens per second for one user on RTX 3060 12 GB in Q4 (estimate based on memory bandwidth).

On which hardware?

Hardware	Cards	Speed	Machine
RTX 3060 12 GB	1 is enough	~22 tok/s	€1,180
RTX 4060 Ti 16 GB	1 is enough	~18 tok/s	€1,350
Radeon RX 7800 XT 16 GB	1 is enough	~38 tok/s	€1,400
RTX 4070 12 GB	1 is enough	~31 tok/s	€1,450
RTX 3090 24 GB (used)	1 is enough	~57 tok/s	€1,650
RTX 4070 Ti SUPER 16 GB	1 is enough	~41 tok/s	€1,750
Radeon RX 7900 XTX 24 GB	1 is enough	~59 tok/s	€1,850
RTX 4080 SUPER 16 GB	1 is enough	~45 tok/s	€1,950
RTX 5080 16 GB	1 is enough	~59 tok/s	€2,100
Mini-PC Ryzen AI Max+ 395 128 GB	1 is enough	~16 tok/s	€2,300
RTX 4090 24 GB	1 is enough	~62 tok/s	€2,900
RTX 5090 32 GB	1 is enough	~110 tok/s	€3,500
NVIDIA DGX Spark 128 GB	1 is enough	~17 tok/s	€4,000
Mac Studio M4 Max 128 GB	1 is enough	~33 tok/s	€4,400
RTX 6000 Ada 48 GB	1 is enough	~59 tok/s	€7,900
RTX PRO 6000 Blackwell 96 GB	1 is enough	~110 tok/s	€9,900
Mac Studio M3 Ultra 512 GB	1 is enough	~50 tok/s	€11,500
L40S 48 GB	1 is enough	~53 tok/s	€13,500
A100 80 GB	1 is enough	~125 tok/s	€17,000
H100 80 GB	1 is enough	~205 tok/s	€34,000
H200 141 GB	1 is enough	~294 tok/s	€39,000
B200 192 GB	1 is enough	~491 tok/s	€48,000

Estimated complete machine: cards + host PC (€900 for 1–2 consumer cards, €2,800 beyond, a €6,000 server for datacenter cards). Macs and mini-PCs are sold complete.