How much memory does Phi-4 need at Q4?

The Q4 figure used here is 10.6 GB. GGUF file size plus an FP16 KV-cache estimate at 8,192 tokens. The cache counts only the layers and heads named in the method on the About page.

What is the cheapest catalog device that runs Phi-4 at Q4?

GeForce RTX 3060 12GB, list price $329.

How fast is Phi-4 on the GeForce RTX 3060 12GB, as a bandwidth estimate?

About 40.5 tok/s at Q4. That is device bandwidth divided by active weight bytes, not a measured run.

Tight. Phi-4 just fits on the GeForce RTX 3060 12GB at Q4, with 0.8 GB of headroom. Usable memory is 11.4 GB.

Can the GeForce RTX 3060 12GB run Phi-4?

Data updated 16 October 2026. Last reviewed the same day, when this catalog snapshot was written.

Usable memory on this 12 GB VRAM config is 11.4 GB (95% of the sticker figure). A result under 2 GB of headroom is labeled tight.

Quant ladder

Estimated memory for Phi-4 on GeForce RTX 3060 12GB
Quant	Estimate	Headroom	Est. tok/s	Verdict
Q4	10.6 GB	0.8 GB	40.5 tok/s	Tight
Q8	17.3 GB	-5.9 GB	23.1 tok/s	No
FP16	35.2 GB	-23.8 GB	12.3 tok/s	No

GGUF file size plus an FP16 KV-cache estimate at 8,192 tokens. The cache counts only the layers and heads named in the method on the About page.
Formula estimate: parameter count × bytes per weight × 1.20. No GGUF file size is stored for this quant.
Tokens/sec is a bandwidth roofline: GB/s divided by active weight GB. It is not a measured generation speed.

KV cache at 8K and 32K

FP16 KV-cache estimate: 1.7 GB at 8,192 tokens and 6.7 GB at 32,768 tokens. GQA counts two tensors (K and V) for the stored KV heads and head dimension.

Offload

The Q4 weight estimate fits in usable VRAM, so this page does not assume a system-RAM offload.

Ollama

This command pulls the Q4_K_M file from unsloth/phi-4-GGUF, the Hub repo whose file size is used above. It is not an official Ollama library tag.

ollama run hf.co/unsloth/phi-4-GGUF:Q4_K_M

Other picks

Cheapest catalog device at Q4: GeForce RTX 3060 12GB
Closest catalog model that still fits at Q4: Qwen2.5 14B Instruct.

All hardware for Phi-4 · All models for the GeForce RTX 3060 12GB

Questions

How much memory does Phi-4 need at Q4?: The Q4 figure used here is 10.6 GB. GGUF file size plus an FP16 KV-cache estimate at 8,192 tokens. The cache counts only the layers and heads named in the method on the About page.
Can the GeForce RTX 3060 12GB run Phi-4?: Tight. Phi-4 just fits on the GeForce RTX 3060 12GB at Q4, with 0.8 GB of headroom. Usable memory is 11.4 GB.
What is the cheapest catalog device that runs Phi-4 at Q4?: GeForce RTX 3060 12GB, list price $329.
How fast is Phi-4 on the GeForce RTX 3060 12GB, as a bandwidth estimate?: About 40.5 tok/s at Q4. That is device bandwidth divided by active weight bytes, not a measured run.