Hardware / NVIDIA /RTX PRO 6000 Blackwell
NVIDIA · GPU
RTX PRO 6000 Blackwell
Runs 144 models fully in memory, 2 more with CPU offload. 2 of those sit close enough to 96 GB that our size estimate is what decides them.
Memory
96 GB
Bandwidth
1792 GB/s
Runs
144
01
Specifications
Memory
96 GB
Bandwidth
1792 GB/s
FP16 compute
125 TFLOPS
TDP
600 W
MSRP
$8,565
Street price
—
Released
Apr 23, 2025
Architecture
Blackwell
02
What it runs
Granite 4.1 8B8.8BQ8_0131K153 tok/sestFits in memoryQwen3.6 35B A3B36BQ4_K_M262K670 tok/sestFits in memoryGemma 4 26B A4B26.5BQ8_0131K287 tok/sestFits in memoryQwen3.5-9B9.7BQ4_K_M262K244 tok/sestFits in memoryQwen3.5-122B-A10B125BQ4_K_M66K201 tok/sestFits in memoryNemotron 3 Nano 30B A3B31.6BQ5_K_M262K571 tok/sestFits in memoryMinistral 3 14B 251213.9BQ4_K_M262K170 tok/sestFits in memoryMinistral 3 8B 25128.9BQ5_K_M262K227 tok/sestFits in memoryMinistral 3 3B 25123.8BQ8_0131K355 tok/sestFits in memoryGranite 4.0 Micro3.2BQ5_K_M66K535 tok/sestFits in memoryQwen3 VL 8B Thinking8.8BQ8_0131K153 tok/sestFits in memoryQwen3 VL 30B A3B Thinking31.1BQ8_0262K382 tok/sestFits in memoryQwen3 Next 80B A3B Thinking81.3BQ4_K_M262K670 tok/sestFits in memoryUI-TARS 7B8.3BQ8_066K163 tok/sestFits in memoryHunyuan A13B Instruct80.4BQ5_K_M131K132 tok/sestFits in memoryLlama Guard 4 12B12BQ4_K_M262K197 tok/sestFits in memoryQwen3 14B14.8BQ4_K_M131K160 tok/sestFits in memoryGemma 3 12B12.2BQ4_K_M131K194 tok/sestFits in memoryRocinante 12B12.2BQ5_K_M66K165 tok/sestFits in memoryLlama 3.2 3B Instruct3.2BQ4_K_M131K739 tok/sestFits in memoryMistral Nemo12.2BQ5_K_M131K165 tok/sestFits in memoryQwen3.5-35B-A3B36BQ5_K_M262K571 tok/sestFits in memoryQwen3 VL 8B Instruct8.8BQ4_K_M262K269 tok/sestFits in memoryQwen3 VL 30B A3B Instruct31.1BQ5_K_M262K571 tok/sestFits in memoryQwen3 Next 80B A3B Instruct81.3BQ4_K_M262K670 tok/sestFits in memoryQwen3 30B A3B Thinking 250730.5BQ4_K_M66K670 tok/sestFits in memoryQwen3 Coder 30B A3B Instruct30.5BQ8_0262K382 tok/sestFits in memoryQwen3 30B A3B Instruct 250730.5BQ4_K_M262K670 tok/sestFits in memoryQwen3 8B8.2BQ4_K_M131K288 tok/sestFits in memoryGemma 3 4B4.3BQ8_0131K314 tok/sestFits in memoryLlama 3.1 8B Instruct8BQ8_0131K169 tok/sestFits in memoryNiagara 19m Batch.en20MQ8_0~262K67464 tok/sestFits in memoryNiagara 38m Batch.en38MQ5_K_M~262K53042 tok/sestFits in memoryARK ASR 0.6B1.3BQ8_0~262K1038 tok/sestFits in memoryARK ASR 3B4.1BQ8_0~262K329 tok/sestFits in memoryAudio8 ASR 0.1B0.3BQ5_K_M~262K6719 tok/sestFits in memoryHiggs Audio v3 8b STT v28.9BQ8_0~262K152 tok/sestFits in memoryHiggs Audio v3 STT2.7BQ8_0~262K500 tok/sestFits in memoryCohere Transcribe 03 20262.1BQ4_K_M~262K1126 tok/sestFits in memoryDistil Large v3.50.8BQ4_K_M~262K2956 tok/sestFits in memoryLite Whisper Large v3 Acc1.4BQ8_0~262K964 tok/sestFits in memoryOwsm CTC v3.1 1B1.1BQ8_0~262K1227 tok/sestFits in memoryOwsm CTC v3.2 ft 1B1BQ4_K_M~262K2365 tok/sestFits in memoryOwsm CTC v4 1B1BQ4_K_M~262K2365 tok/sestFits in memoryHubert Large Ls960 ft0.3BQ4_K_M~262K7390 tok/sestFits in memoryMMS 1b All1BQ8_0~262K1349 tok/sestFits in memoryWav2vec2 Large 960h Lv60 Self0.3BQ8_0~262K4217 tok/sestFits in memoryHojo ASR V15.2BQ8_0~262K261 tok/sestFits in memoryGranite 4.0 1b Speech2.3BQ4_K_M~262K1028 tok/sestFits in memoryGranite Speech 3.3 2b3BQ5_K_M~262K672 tok/sestFits in memoryGranite Speech 3.3 8b8.6BQ4_K_M~262K275 tok/sestFits in memoryGranite Speech 4.1 2b2.3BQ4_K_M~262K1028 tok/sestFits in memoryGranite Speech 4.1 2b NAR2.3BQ5_K_M~262K876 tok/sestFits in memorySTT 2.6b en2.6BQ5_K_M~262K775 tok/sestFits in memoryPhi 4 Multimodal Instruct5.6BQ8_0~262K241 tok/sestFits in memoryVibeVoice ASR HF8.3BQ8_0~262K163 tok/sestFits in memoryVoxtral Mini 3B 25075BQ4_K_M~262K473 tok/sestFits in memoryVoxtral Mini 4B Realtime 26024.4BQ4_K_M~262K537 tok/sestFits in memoryCanary 180m Flash0.2BQ4_K_M~262K13137 tok/sestFits in memoryCanary 1b1BQ5_K_M~262K2016 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.