RunsLocalGPU checkerLocal AI toolsv1.8 · data Oct 14, 2026

All models / Phi-4 14B

What GPU do I need to run Phi-4 14B?

Phi-4 14B has 14.7B parameters. At Q4_K_M with 8K context it needs about 11.0 GB of memory. The smallest graphics card that runs it fully on the GPU is about 12 GB, such as the RTX 3060 12GB.

Check your own hardware

Memory by quantization (8K context)

QuantBits per weightMemory needed
Q8_08.517.7 GB
Q6_K6.5614.2 GB
Q5_K_M5.6912.6 GB
Q4_K_M4.8511.0 GB
Q3_K_M3.919.3 GB
IQ2_M2.77.1 GB

Hardware that runs it fully in fast memory

HardwareMemoryBest quantEst. speed
RTX 3080 Ti12 GBQ4_K_M61 tok/s
RTX 3080 12GB12 GBQ4_K_M61 tok/s
RTX 507012 GBQ4_K_M45 tok/s
RTX 4070 Ti12 GBQ4_K_M34 tok/s
RTX 4070 Super12 GBQ4_K_M34 tok/s
RTX 407012 GBQ4_K_M34 tok/s
Arc B58012 GBQ4_K_M31 tok/s
RX 7700 XT12 GBQ4_K_M29 tok/s
RTX 3060 12GB12 GBQ4_K_M24 tok/s
RTX 508016 GBQ6_K48 tok/s
RTX 5070 Ti16 GBQ6_K45 tok/s
RTX 4080 Super16 GBQ6_K37 tok/s
RTX 408016 GBQ6_K36 tok/s
RTX 4070 Ti Super16 GBQ6_K33 tok/s
RX 9070 XT16 GBQ6_K32 tok/s
RX 907016 GBQ6_K32 tok/s
RX 7800 XT16 GBQ6_K31 tok/s
Arc A770 16GB16 GBQ6_K28 tok/s
RX 6800 XT16 GBQ6_K25 tok/s
RTX 5060 Ti 16GB16 GBQ6_K22 tok/s
RX 9060 XT 16GB16 GBQ6_K16 tok/s
RTX 4060 Ti 16GB16 GBQ6_K14 tok/s
M5 16GB16 GBQ4_K_M10 tok/s
M4 16GB16 GBQ4_K_M8.1 tok/s
RX 7900 XT20 GBQ8_031 tok/s
RTX 409024 GBQ8_039 tok/s
RTX 3090 Ti24 GBQ8_039 tok/s
RX 7900 XTX24 GBQ8_037 tok/s
RTX 309024 GBQ8_036 tok/s
M4 Pro 24GB24 GBQ6_K14 tok/s
RTX 509032 GBQ8_069 tok/s
M5 32GB32 GBQ8_05.9 tok/s
M4 32GB32 GBQ8_04.6 tok/s
M4 Max 36GB36 GBQ8_016 tok/s
M4 Pro 48GB48 GBQ8_010 tok/s
M4 Max 64GB64 GBQ8_021 tok/s
Ryzen AI Max+ 395 64GB64 GBQ8_09.8 tok/s
RTX PRO 6000 Blackwell96 GBQ8_069 tok/s
M3 Ultra 96GB96 GBQ8_031 tok/s
M4 Max 128GB128 GBQ8_021 tok/s
DGX Spark 128GB128 GBQ8_010 tok/s
Ryzen AI Max+ 395 128GB128 GBQ8_09.8 tok/s
M3 Ultra 256GB256 GBQ8_031 tok/s
M3 Ultra 512GB512 GBQ8_031 tok/s

Runs with offload to system RAM (slower)

Assumes 32 GB of DDR5 system RAM for offload. Estimates only; real speeds usually land within about 20%.