~/best-gpu-for/gpt-oss-20b
OpenAI

Best GPU for gpt-oss 20B

Ranked recommendations from the 42-GPU catalog for running gpt-oss 20B (20.9B params). Top picks by quality, then by cost.

Top picks.

$ ./rank --by quality
Tightest budget
RTX 4060 Ti 16GB
16GB · runs at Q4_K_M
12 GB weights · 0.39 GB headroom
Best quality
M2 Max 64
64GB · 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 gpt-oss-20b
GPUVRAMBest quantWeights at 15% safetyFit
appleM2 Max 6464GBFP16/BF16
42fitsappleM3 Max 6464GBFP16/BF16
42fitsappleM4 Pro 6464GBFP16/BF16
42fitsappleM5 Pro 6464GBFP16/BF16
42fitsnvidiaH100 80GB80GBFP16/BF16
42fitsappleM3 Max 9696GBFP16/BF16
42fitsnvidiaRTX Pro 600096GBFP16/BF16
42fitsappleM4 Max 128128GBFP16/BF16
42fitsappleM5 Max 128128GBFP16/BF16
42fitsnvidiaDGX Spark128GBFP16/BF16
42fitsintelGaudi 3128GBFP16/BF16
42fitsnvidiaH200141GBFP16/BF16
42fitsnvidia2× H100 NVL188GBFP16/BF16
42fitsappleM2 Ultra 192192GBFP16/BF16
42fitsnvidiaB200192GBFP16/BF16
42fitsamdMI300X192GBFP16/BF16
42fitsappleM3 Ultra 512512GBFP16/BF16
42fitsnvidiaRTX 509032GBQ8_0
22fits

Next steps.

$ ./next
$ checkCan I run gpt-oss 20B?see all quants →full quant matrix$ tuneTune the mathopen calculator →/calc?model=gpt-oss-20b

Discussion.

$ gh discussion list

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