can-i-run-this-llm

Can I run this LLM?

Can a NVIDIA RTX 3060 (12GB) run Phi-4 14B?

Phi · MIT · estimated for a 12.0 GB machine at 4K context.

Yes — it runs great.

Best fit: Phi-4 14B at Q3_K_M · 7.2 GB · 30–46 tok/s on a NVIDIA RTX 3060 (12GB).

Every quantization on this machine

QuantSizeSpeedFit
Q2_K5.6 GB37–56 tok/sRuns great
Q3_K_M7.2 GB30–46 tok/sRuns great
Q4_K_M8.9 GB25–38 tok/sTight / slow
Q5_K_M10.4 GB22–33 tok/sWon't fit
Q6_K12.0 GB20–29 tok/sWon't fit
Q8_015.6 GB16–23 tok/sWon't fit
FP1629.3 GB9–13 tok/sWon't fit

Ran Phi-4 14B on a NVIDIA RTX 3060 (12GB)? Add your real numbers →