Hardware / NVIDIA /H100 80GB SXM
NVIDIA · GPU

H100 80GB SXM

Runs 136 models fully in memory, 8 more with CPU offload. 4 of those sit close enough to 80 GB that our size estimate is what decides them.

Memory
80 GB
Bandwidth
3350 GB/s
Runs
136
01

Specifications

Memory
80 GB
Bandwidth
3350 GB/s
FP16 compute
989 TFLOPS
TDP
700 W
MSRP
$30,000
Street price
Released
Oct 1, 2022
Architecture
Hopper
02

What it runs

Granite 4.1 8B8.8BQ8_0131K287 tok/sestFits in memoryQwen3.6 35B A3B36BQ4_K_M262K1253 tok/sestFits in memoryGemma 4 26B A4B26.5BQ8_0131K536 tok/sestFits in memoryQwen3.5-9B9.7BQ5_K_M262K389 tok/sestFits in memoryQwen3.5-27B27.8BQ5_K_M131K136 tok/sestFits in memoryGLM 4.7 Flash31.2BQ4_K_M131K120 tok/sestFits in memoryNemotron 3 Nano 30B A3B31.6BQ5_K_M262K1068 tok/sestFits in memoryMinistral 3 14B 251213.9BQ4_K_M262K318 tok/sestFits in memoryMinistral 3 8B 25128.9BQ5_K_M262K423 tok/sestFits in memoryMinistral 3 3B 25123.8BQ4_K_M131K1163 tok/sestFits in memoryOlmo 3 32B Think32.2BQ4_K_M66K137 tok/sestFits in memorygpt-oss-safeguard-20b21.5BQ5_K_M131K149 tok/sestFits in memoryQwen3 VL 32B Instruct33.4BQ4_K_M131K132 tok/sestFits in memoryGranite 4.0 Micro3.2BQ4_K_M66K1174 tok/sestFits in memoryQwen3 VL 8B Thinking8.8BQ8_0131K287 tok/sestFits in memoryQwen3 VL 30B A3B Thinking31.1BQ8_0262K715 tok/sestFits in memoryCydonia 24B V4.123.6BQ4_K_M131K187 tok/sestFits in memoryQwen3 Next 80B A3B Thinking81.3BQ4_K_M131K1253 tok/sestFits in memoryUI-TARS 7B8.3BQ8_066K304 tok/sestFits in memoryUncensored24BQ4_K_M66K184 tok/sestFits in memoryHunyuan A13B Instruct80.4BQ4_K_M131K289 tok/sestFits in memoryMistral Small 3.2 24B24BQ5_K_M131K157 tok/sestFits in memoryLlama Guard 4 12B12BQ4_K_M262K368 tok/sestFits in memoryQwen3 14B14.8BQ8_0131K170 tok/sestFits in memoryGemma 3 12B12.2BQ5_K_M131K309 tok/sestFits in memoryGemma 3 27B27.4BQ4_K_M262K161 tok/sestFits in memoryRocinante 12B12.2BQ5_K_M66K309 tok/sestFits in memoryLlama 3.2 3B Instruct3.2BQ8_0131K788 tok/sestFits in memoryMistral Nemo12.2BQ5_K_M131K309 tok/sestFits in memoryQwen3.6 27B27.8BQ4_K_M131K159 tok/sestFits in memoryQwen3.5-35B-A3B36BQ4_K_M262K1253 tok/sestFits in memoryQwen3 VL 8B Instruct8.8BQ4_K_M262K502 tok/sestFits in memoryQwen3 VL 30B A3B Instruct31.1BQ5_K_M262K1068 tok/sestFits in memoryQwen3 Next 80B A3B Instruct81.3BQ5_K_M131K1068 tok/sestFits in memoryQwen3 30B A3B Thinking 250730.5BQ5_K_M66K1068 tok/sestFits in memoryQwen3 Coder 30B A3B Instruct30.5BQ4_K_M262K1253 tok/sestFits in memoryQwen3 30B A3B Instruct 250730.5BQ8_0262K715 tok/sestFits in memoryQwen3 8B8.2BQ8_0131K308 tok/sestFits in memoryQwen3 32B32.8BQ4_K_M131K135 tok/sestFits in memoryMistral Small 3.1 24B24BQ5_K_M66K157 tok/sestFits in memoryGemma 3 4B4.3BQ8_0131K587 tok/sestFits in memoryReka Flash 320.9BQ8_066K121 tok/sestFits in memoryLlama 3.1 8B Instruct8BQ8_0131K315 tok/sestFits in memoryNiagara 19m Batch.en20MQ8_0~262K126118 tok/sestFits in memoryNiagara 38m Batch.en38MQ4_K_M~262K116332 tok/sestFits in memoryARK ASR 0.6B1.3BQ5_K_M~262K2899 tok/sestFits in memoryARK ASR 3B4.1BQ8_0~262K615 tok/sestFits in memoryAudio8 ASR 0.1B0.3BQ5_K_M~262K12560 tok/sestFits in memoryHiggs Audio v3 8b STT v28.9BQ4_K_M~262K497 tok/sestFits in memoryHiggs Audio v3 STT2.7BQ5_K_M~262K1396 tok/sestFits in memoryCohere Transcribe 03 20262.1BQ5_K_M~262K1794 tok/sestFits in memoryDistil Large v3.50.8BQ8_0~262K3153 tok/sestFits in memoryLite Whisper Large v3 Acc1.4BQ8_0~262K1802 tok/sestFits in memoryOwsm CTC v3.1 1B1.1BQ4_K_M~262K4019 tok/sestFits in memoryOwsm CTC v3.2 ft 1B1BQ8_0~262K2522 tok/sestFits in memoryOwsm CTC v4 1B1BQ8_0~262K2522 tok/sestFits in memoryHubert Large Ls960 ft0.3BQ8_0~262K7882 tok/sestFits in memoryMMS 1b All1BQ5_K_M~262K3768 tok/sestFits in memoryWav2vec2 Large 960h Lv60 Self0.3BQ4_K_M~262K13814 tok/sestFits in memoryHojo ASR V15.2BQ4_K_M~262K853 tok/sestFits in memory

A speed marked est is calculated from this card's memory bandwidth and the size of the weights it has to read per word. It is not a measurement.

Something wrong on this page? Tell us