OpenAI on NVIDIA Hopper

Can I run gpt-oss 120B-A5.1B on an H100 SXM 80GB?

Yes. gpt-oss 120B-A5.1B at Q4_K_M uses 67.4 GB of the 74.4 GB available on a H100 SXM 80GB, and runs at about 323 tokens per second. You can push the context to 108K.

Fits, but tight

67.4 GB of 74.4 GB · 91%
074 GB
Weights 66.0 GB
KV cache 0.6 GB
Runtime overhead 0.8 GB

This fits with almost nothing to spare. A background application claiming VRAM will push it over. Drop to the next quantisation down, or quantise the KV cache to Q8_0 — that halves the cache for no meaningful quality loss.

Generation323tok/s
Prompt processing40724tok/s
Max context108Ktokens
KV per 1K tokens0GB

Every quantisation of gpt-oss 120B-A5.1B on a H100 SXM 80GB

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 217.6 GB 218.9 GB Won't fit — 6.71
INT8 / W8A8 8.50 115.3 GB 116.7 GB Won't fit — 20.2
FP8 (E4M3) 8.00 108.8 GB 110.1 GB Won't fit — 24.6
Q6_K 6.56 89.2 GB 90.6 GB Won't fit — 46.8
Q5_K_M 5.67 77.1 GB 78.5 GB Won't fit — 129
Q5_K_S 5.52 75.1 GB 76.4 GB Won't fit — 183
Q4_K_M 4.85 66.0 GB 67.4 GB Fits, but tight 108K 323
Q4_K_S 4.58 62.4 GB 63.8 GB Runs comfortably 128K 331
Q4_0 4.55 62.0 GB 63.4 GB Runs comfortably 128K 332
AWQ 4-bit 4.25 59.4 GB 60.7 GB Runs comfortably 128K 338
GPTQ 4-bit 4.25 59.4 GB 60.7 GB Runs comfortably 128K 338
MXFP4 4.25 59.4 GB 60.7 GB Runs comfortably 128K 338
IQ4_XS 4.25 58.0 GB 59.3 GB Runs comfortably 128K 342
Q3_K_M 3.91 53.4 GB 54.7 GB Runs comfortably 128K 354
IQ3_M 3.70 50.6 GB 51.9 GB Runs comfortably 128K 362
IQ3_XXS 3.06 41.9 GB 43.3 GB Runs comfortably 128K 388
Q2_K 2.63 36.1 GB 37.5 GB Runs comfortably 128K 408
IQ2_XXS 2.06 28.5 GB 29.8 GB Runs comfortably 128K 438
IQ1_M 1.75 24.3 GB 25.7 GB Runs comfortably 128K 456

Also worth checking

gpt-oss 120B-A5.1B on every GPU
Full hardware tableEverything that runs on a H100 SXM 80GB
Full model tableWhy the KV cache is the part that bites
Guide