NVIDIA · workstation

RTX PRO 5000 Blackwell

RTX PRO 5000 Blackwell has 72 GB of VRAM at 1344 GB/s — about 66.96 GiB usable after driver and compositor overhead. 2071 of 2118 indexed models fit at 16K context with q4_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
72 GB
GDDR7
Bandwidth
1344 GB/s
384-bit bus
Tensor FP16
295 TF
dense
TDP
300 W
$4569 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1781vision language 186image 2audio asr 39audio tts 21video 16embedding 26

What fits at 16K context

largest quantization that fits, per model · 2071 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
CalmeRys-78B-Orpo-v0.1Q6_K78.0B64.27 GiB1.51 GiB66.92 GiB0.04 GiB12±22%
calme-2.3-rys-78bQ6_K78.0B64.27 GiB1.51 GiB66.92 GiB0.04 GiB12±22%
Step-3.5-Flash-REAP-121B-A11BI1-Q4_K_S121B63.83 GiB1.98 GiB66.84 GiB0.12 GiB12±22%
Qwen3-VL-235B-A22B-ThinkingMoEUD-IQ1_M236B64.90 GiB0.83 GiB66.76 GiB0.20 GiB50±37%
Qwen3.5-99BMoEI1-Q5_K_M99.0B65.60 GiB0.11 GiB66.74 GiB0.22 GiB66±37%
Step-3.7-FlashIQ2_M201B63.68 GiB1.98 GiB66.69 GiB0.27 GiB12±22%
Qwen3-VL-235B-A22B-InstructMoEUD-IQ1_M236B64.83 GiB0.83 GiB66.69 GiB0.27 GiB50±37%
GLM-4.5-Air-DerestrictedMoEQ4_1110B64.77 GiB0.81 GiB66.61 GiB0.35 GiB50±37%
GLM-4.5-AirMoEQ4_1110B64.77 GiB0.81 GiB66.61 GiB0.35 GiB50±37%
MiniMax-M2.7MoEIQ2_S229B64.38 GiB1.09 GiB66.46 GiB0.50 GiB62±37%
Ornith-Agents-A1-3.7-35B-A3B-dare_ties_v4MoEF1634.7B65.34 GiB0.09 GiB66.43 GiB0.53 GiB71±37%
Ornith-Agents-A1-3.6-35B-A3B-dare_tiesMoEF1634.7B65.34 GiB0.09 GiB66.43 GiB0.53 GiB71±37%
DeepSeek-Coder-V2-Instruct-0724MoEIQ2_S236B65.07 GiB0.30 GiB66.40 GiB0.56 GiB65±37%
DeepSeek-V2.5MoEIQ2_S236B65.07 GiB0.30 GiB66.40 GiB0.56 GiB65±37%
DeepSeek-Coder-V2-InstructMoEIQ2_S236B65.07 GiB0.30 GiB66.40 GiB0.56 GiB65±37%
command-a-plus-05-2026-bf16MoEIQ2_S219B65.10 GiB0.26 GiB66.37 GiB0.59 GiB54±37%
dots.llm1.instMoEUD-IQ3_XXS143B60.94 GiB4.36 GiB66.33 GiB0.63 GiB40±37%
Llama-4-Scout-17B-16E-InstructMoEKV unresolvedQ4_1109B64.35 GiB0.84 GiB66.22 GiB0.74 GiB51±37%
MiMo-V2-FlashMoEKV unresolvedI1-IQ1_M310B64.64 GiB0.53 GiB66.22 GiB0.74 GiB67±37%
Mistral-Small-4-119B-2603MoEQ4_K_S119B65.08 GiB0.10 GiB66.22 GiB0.74 GiB72±37%
MiniMax-M2MoEUD-IQ1_M229B64.01 GiB1.09 GiB66.09 GiB0.87 GiB63±37%
grok-2MoEIQ2_XXS270B63.81 GiB1.13 GiB66.08 GiB0.88 GiB22±37%
Qwen3.5-122B-A10B-hereticMoEI1-Q4_K_S123B64.86 GiB0.11 GiB65.99 GiB0.97 GiB72±37%
Qwen3-235B-A22B-abliteratedMoEI1-IQ2_XS235B64.09 GiB0.83 GiB65.95 GiB1.01 GiB51±37%
Qwen3.5-REAP-212B-A17BMoEIQ2_M212B64.76 GiB0.13 GiB65.95 GiB1.01 GiB67±37%
MiniMax-M2.1MoEUD-IQ1_M229B63.74 GiB1.09 GiB65.81 GiB1.15 GiB63±37%
MiniMax-M2.5MoEUD-IQ1_M229B63.74 GiB1.09 GiB65.81 GiB1.15 GiB63±37%
NVIDIA-Nemotron-3-Super-120B-A12B-BF16MoEIQ4_NL124B64.39 GiB0.39 GiB65.77 GiB1.19 GiB62±37%
XORTRON-NXTXPRTXXLIQ4_XS128B63.03 GiB1.55 GiB65.74 GiB1.22 GiB12±22%
Qwen3.6-35B-A3B-uncensored-hereticMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Darwin-35B-A3B-OpusMoEBF1636.0B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Carnice-MoE-35B-A3BMoEF1636.0B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Qwen3.6-35B-A3B-hereticMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Aurora-Code-1MoEBF1634.7B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
grug-35b-v2MoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
grug-35bMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
WorldSim-Opus-3.6-35B-A3BMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Qwen3.6-35B-A3B-abliterated-MAXMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Huihui-Nex-N2-mini-abliteratedMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Qwen3.6-35B-A3B-AnkoMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
KAT-Coder-V2.5-DevMoEBF1634.7B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Ornith-1.0-35B-uncensored-hereticMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Qwen3.6-35B-A3B-abliterated-v4MoEBF1634.7B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Qwen3.5-35B-A3B-ultra-uncensored-hereticMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Nex-N2-mini-ultra-uncensored-hereticMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Agents-A1MoEF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Qwen3.6-35B-A3B-java-v1MoEBF1634.7B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Nex-N2-miniMoEBF1635.1B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Qwen3.5-35B-A3B-BaseMoEBF1636.0B64.61 GiB0.09 GiB65.70 GiB1.26 GiB72±37%
Qwen3.5-REAP-262B-A17BMoEIQ2_XXS262B64.50 GiB0.13 GiB65.69 GiB1.27 GiB73±37%
Laguna-S-2.1MoEQ4_K_S118B64.36 GiB0.25 GiB65.64 GiB1.32 GiB66±37%
MiniMax-M2.7-BF16-ultra-uncensored-hereticMoEI1-IQ2_S229B63.36 GiB1.09 GiB65.44 GiB1.52 GiB63±37%
GLM-4.7-REAP-218B-A32BMoEUD-IQ1_M218B62.63 GiB1.62 GiB65.29 GiB1.67 GiB42±37%
Devstral-2-123B-Instruct-2512IQ4_XS125B62.51 GiB1.55 GiB65.22 GiB1.74 GiB12±22%
Mixtral-8x22B-Instruct-v0.1MoEQ3_K_M141B63.14 GiB0.98 GiB65.18 GiB1.78 GiB22±37%
Mixtral-8x22B-v0.1MoEQ3_K_M141B63.14 GiB0.98 GiB65.18 GiB1.78 GiB22±37%
Mixtral-8x22B-v0.1MoEQ3_K_M141B63.13 GiB0.98 GiB65.18 GiB1.78 GiB22±37%
Mistral-Medium-3.5-128BI1-IQ4_XS128B62.47 GiB1.55 GiB65.17 GiB1.79 GiB12±22%
Qwen3.5-122B-A10BMoEIQ4_XS125B63.77 GiB0.11 GiB64.90 GiB2.06 GiB73±37%
HunyuanImage-2.1Q5_117.5B63.57 GiB0.00 GiB64.62 GiB2.34 GiB12±22%
●From the file◐Predictedwhat these mean

Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.

Questions people ask

What AI models can a RTX PRO 5000 Blackwell run?
2071 of 2118 indexed open-weight models fit a RTX PRO 5000 Blackwell at 16,384 context with q4_0 KV cache, the largest being CalmeRys-78B-Orpo-v0.1 at Q6_K. That covers text, vision-language, image, video and speech models.
How much usable memory does a RTX PRO 5000 Blackwell actually have?
Its nameplate is 72 GB, but about 66.96 GiB is available to a model once driver and compositor overhead is accounted for.
Is a RTX PRO 5000 Blackwell fast for local AI?
Its memory bandwidth is 1344 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.
RTX PRO 5000 Blackwell — what AI models can it run locally? — ossmodeldb