~/best-gpu-for/deepseek-r1
DeepSeek

Best GPU for DeepSeek R1

Ranked recommendations from the 42-GPU catalog for running DeepSeek R1 (671B params). Top picks by quality, then by cost.

Top picks.

$ ./rank --by quality
Tightest budget
M3 Ultra 512
512GB · runs at Q4_K_M
377 GB weights · 55 GB headroom
Best quality
M3 Ultra 512
512GB · runs at Q4_K_M
smallest card supporting the best available quant
Apple Silicon
M3 Ultra 512
512GB · runs at Q4_K_M
smallest M-series that fits

Full ranked list.

$ ./vrambudget --rank deepseek-r1
GPUVRAMBest quantWeights at 15% safetyFit
appleM3 Ultra 512512GBQ4_K_M
377fitsnvidiaRTX 30708GBQ4_K_M
377overnvidiaRTX 40608GBQ4_K_M
377overnvidiaRTX 308010GBQ4_K_M
377overnvidiaRTX 3060 12GB12GBQ4_K_M
377overnvidiaRTX 3080 Ti12GBQ4_K_M
377overnvidiaRTX 407012GBQ4_K_M
377overnvidiaRTX 4070 Super12GBQ4_K_M
377overnvidiaRTX 507012GBQ4_K_M
377overintelArc B58012GBQ4_K_M
377overnvidiaRTX 4060 Ti 16GB16GBQ4_K_M
377overnvidiaRTX 4070 Ti Super16GBQ4_K_M
377overnvidiaRTX 408016GBQ4_K_M
377overnvidiaRTX 4080 Super16GBQ4_K_M
377overnvidiaRTX 5070 Ti16GBQ4_K_M
377overnvidiaRTX 508016GBQ4_K_M
377overnvidiaRTX 309024GBQ4_K_M
377overnvidiaRTX 3090 Ti24GBQ4_K_M
377over

Next steps.

$ ./next
$ checkCan I run DeepSeek R1?see all quants →full quant matrix$ tuneTune the mathopen calculator →/calc?model=deepseek-r1

Discussion.

$ gh discussion list

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