~/best-gpu-for/phi-4
Microsoft

Best GPU for Phi-4

Ranked recommendations from the 42-GPU catalog for running Phi-4 (14.7B params). Top picks by quality, then by cost.

Top picks.

$ ./rank --by quality
Tightest budget
RTX 3080
10GB · runs at Q3_K_M
6.3 GB weights · 0.88 GB headroom
Best quality
RTX A6000
48GB · runs at FP16/BF16
smallest card supporting the best available quant
Apple Silicon
M2 Max 64
64GB · runs at FP16/BF16
smallest M-series that fits
Datacenter (FP16 ambition)
H100 80GB
80GB · runs at FP16/BF16
best-quality datacenter card that fits

Full ranked list.

$ ./vrambudget --rank phi-4
GPUVRAMBest quantWeights at 15% safetyFit
nvidiaRTX A600048GBFP16/BF16
29fitsnvidiaRTX 6000 Ada48GBFP16/BF16
29fitsnvidiaL40S48GBFP16/BF16
29fitsamdRadeon Pro W790048GBFP16/BF16
29fitsappleM2 Max 6464GBFP16/BF16
29fitsappleM3 Max 6464GBFP16/BF16
29fitsappleM4 Pro 6464GBFP16/BF16
29fitsappleM5 Pro 6464GBFP16/BF16
29fitsnvidiaH100 80GB80GBFP16/BF16
29fitsappleM3 Max 9696GBFP16/BF16
29fitsnvidiaRTX Pro 600096GBFP16/BF16
29fitsappleM4 Max 128128GBFP16/BF16
29fitsappleM5 Max 128128GBFP16/BF16
29fitsnvidiaDGX Spark128GBFP16/BF16
29fitsintelGaudi 3128GBFP16/BF16
29fitsnvidiaH200141GBFP16/BF16
29fitsnvidia2× H100 NVL188GBFP16/BF16
29fitsappleM2 Ultra 192192GBFP16/BF16
29fits

Next steps.

$ ./next
$ checkCan I run Phi-4?see all quants →full quant matrix$ tuneTune the mathopen calculator →/calc?model=phi-4

Discussion.

$ gh discussion list

// sign in with github to leave a comment. threads live in the repo's discussions tab.