NVIDIA RTX PRO 5000 Blackwell

GPU

Blackwell; 48GB VRAM, 1344 GB/s bandwidth, 300W TDP, CUDA cores 14,080. Release date: 2025-03-18. Source: LCZ hardware catalog, with missing fields supplemented from NVIDIA official materials.

9

Run records

3

Models covered

2

Frameworks covered

L0 Self-reported × 9

Evidence mix

Typical measured performance

Published benchmark averages per model × framework × quantization on this hardware.

  1. SGLang · IQ2_XS Decode 4.3 · Prefill 32.3 ·
  2. SGLang · IQ3_S Decode 3.6 · Prefill 27 ·
  3. vLLM · IQ2_XS Decode 4.5 · Prefill 33.8 ·
  4. vLLM · IQ3_S Decode 3.7 · Prefill 27.8 ·
  5. SGLang · IQ2_XS Decode 4.3 · Prefill 32.3 ·
  6. SGLang · IQ3_S Decode 3.6 · Prefill 27 ·
  7. vLLM · IQ2_XS Decode 4.4 · Prefill 33 ·
  8. vLLM · IQ3_S Decode 3.7 · Prefill 27.8 ·
  9. SGLang · FP8 Decode 31.5 · Prefill 5089 ·

Fit check

Physics-based estimate from VRAM and memory bandwidth: pick a model to see which quantizations fit and the expected speed

Formula: tok/s = (bandwidth ÷ model size) × 0.55, adapted from llmfit; see run records for measured numbers

Key specs

Vendor
NVIDIA
Architecture
Blackwell
VRAM
48 GB
Memory bandwidth
1344 GB/s
TDP
300 W
Release date
2025-03-18

Run records

9 configurations

ModelFrameworkQuantizationDecodeGen timeSamples
GLM-5.3SGLangIQ2_XS4.3 tok/s—1
GLM-5.3SGLangIQ3_S3.6 tok/s—1
GLM-5.3vLLMIQ2_XS4.5 tok/s—1
GLM-5.3vLLMIQ3_S3.7 tok/s—1
DeepSeek-V4.1-FlashSGLangIQ2_XS4.3 tok/s—1
DeepSeek-V4.1-FlashSGLangIQ3_S3.6 tok/s—1
DeepSeek-V4.1-FlashvLLMIQ2_XS4.4 tok/s—1
DeepSeek-V4.1-FlashvLLMIQ3_S3.7 tok/s—1
Qwen3.8-27BSGLangFP831.5 tok/s—1