NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition, AI & Machine Learning Benchmarks & Specs

96GB · AI Score 45.3/100 · anchored estimate vs 51 measured cards

45.3 AI Score Includes estimates

We have not run NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition on our bench. These figures are anchored estimates, interpolated per workload against the 51 GPUs we did measure. On Llama 3.1 8B (Q4_K_M) NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition should deliver about 245.5 tokens/sec. Stepping up to Qwen3 32B it should hold roughly 66.6 tok/s. The full Llama 3.3 70B still runs, at about 33.1 tok/s. For image generation, SDXL should run near 10.9 it/s, and FLUX.1-dev at 2.52 it/s. 1 of the 12 workloads won't fit on 96GB at the tested precision, Wan 2.2 5B. We publish those as hard gates rather than quietly dropping to a smaller quant.

AI & Machine Learning benchmark results

Text Generation tok/s 5

Qwen3 4B344.7
Llama 3.1 8B245.5
Qwen2.5-Coder 14B140.4
Qwen3 32B66.6
Llama 3.3 70B33.1
WorkloadResultTelemetryData
Qwen3 4B344.7 tok/sestimatedEst.
Llama 3.1 8B245.5 tok/sestimatedEst.
Qwen2.5-Coder 14B140.4 tok/sestimatedEst.
Qwen3 32B66.6 tok/sestimatedEst.
Llama 3.3 70B33.1 tok/sestimatedEst.

Image Generation images/min 3

Stable Diffusion XL21.8
Z-Image Turbo12.15
FLUX.1 dev5.4
WorkloadResultTelemetryData
Stable Diffusion XL21.8 images/minestimatedEst.
Z-Image Turbo12.15 images/minestimatedEst.
FLUX.1 dev5.4 images/minestimatedEst.

Image Editing images/min 2

WorkloadResultTelemetryData
FLUX.1 Kontext dev2.4 images/minestimatedEst.
Qwen-Image-Edit2.06 images/minestimatedEst.

Video Generation frames/s 2

WorkloadResultTelemetryData
LTX-Video (distilled)12.76 frames/sestimatedEst.
Wan 2.2 5B (720p)n/a frames/sestimatedEst.
How this estimate is derived. This card hasn’t been through our bench yet, so its numbers are anchored estimates, interpolated from the 51 first-party measured cards (Sibling-anchored: same GB202 die, power-capped (~300W); LLM ×0.95 (memory-bound), diffusion/video ×0.78 (power-bound)). The VRAM “won’t fit” gates are exact, since they’re pure capacity limits. Estimates are replaced with measured data as more silicon goes through the bench. Full methodology →

NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition specifications

ArchitectureBlackwell
CUDA cores24,064
VRAM96GB GDDR7 ECC
Memory bus512-bit
Memory bandwidth1792 GB/s
Boost clock2,617 MHz
TDP300 W
Process4nm (TSMC 4N)
InterfacePCIe 5.0 x16
Release date2025-03-18
Launch MSRP$8,565

Verdict, capable, but 96GB sets the ceiling

NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition scores 45.3/100, #12 of 102. It ran 11 of 12; 1 exceeded its 96GB. Figures are anchored estimates, not measurements, we flag that on every row.

Relative performance: where the NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition lands

100% = this card, AI & Machine Learning headline metric (AI Score). #2 of 20 desktop cards in this vertical.

GPURelative%AI Score
NVIDIA RTX PRO 6000 Blackwell Workstation Edition
117%53.1
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition
100%45.3
NVIDIA RTX PRO 5000 Blackwell
70%31.6
NVIDIA RTX A6000
46%21
NVIDIA RTX PRO 4500 Blackwell
28%12.9
AMD Radeon Pro W7900
28%12.8

← All AI & Machine Learning GPU rankings

What this card can build

Whole-job timings, composed from our measured per-model results on this card.

WorkflowTimeEnergyBasis
24-frame storyboard2.3 minn/aestimate, 0 of 2 stages measured
60-second AI short film2.8 minn/aestimate, 0 of 3 stages measured
6-panel comic page3.8 minn/aestimate, 0 of 3 stages measured
Character sheet, 12 poses5.2 minn/aestimate, 0 of 2 stages measured
Full codebase review7.1 minn/aestimate, 0 of 1 stage measured
20 long-form articles14.1 minn/aestimate, 0 of 1 stage measured
40-product photo shoot18.5 minn/aestimate, 0 of 2 stages measured
100-photo restoration batch41.7 minn/aestimate, 0 of 1 stage measured