About

Overview

Performance

120B MoE inference: Up to 130 tokens/s
Token generation at 130 tokens/s:
- 468,000 tokens/hour
- 11.23 million tokens/day
- 337 million tokens/month at continuous utilization
70B model inference: 40–50 tokens/s
405B model inference: 10–15 tokens/s
Prompt processing: Up to 8,400 tokens/s
Concurrent capacity: 40–50 active users
SDXL image generation: 1.25 seconds per 1024×1024 image at 20 steps
Video generation: Production-ready 5-second 720p clips
Total GPU memory: 192 GB GDDR6 ECC

Cost Savings

For teams spending $2,000+ per month on AI and approximately $2,500 per month on Sonnet, Opus, image, and video generation, this workstation moves recurring workloads to controlled, local infrastructure.

Paid off in approximately 16 months
Save approximately $47,124 over three years
Save approximately $104,931 over five years

Characteristics

GPU 2x RTX PRO 6000 Max-Q

CPU Intel Xeon 658X Enterprise (24 Cores)

Board ASUS Pro WS W890E-SAGE SE

VRAM 192 GB GDDR6 Max-Q

System RAM 256GB Kingston DDR5 ECC (64GB x 4)

Storage 2TB Boot NVMe + 2TB Active NVMe SSD

Mass Storage 12TB 7200RPM Enterprise HDD

Power 1700W Pro PSU

Cooling Air-cool GPUs · Liquid-cool CPU

Overview

Performance

120B MoE inference: Up to 130 tokens/s
Token generation at 130 tokens/s:
- 468,000 tokens/hour
- 11.23 million tokens/day
- 337 million tokens/month at continuous utilization
70B model inference: 40–50 tokens/s
405B model inference: 10–15 tokens/s
Prompt processing: Up to 8,400 tokens/s
Concurrent capacity: 40–50 active users
SDXL image generation: 1.25 seconds per 1024×1024 image at 20 steps
Video generation: Production-ready 5-second 720p clips
Total GPU memory: 192 GB GDDR6 ECC

Cost Savings

For teams spending $2,000+ per month on AI and approximately $2,500 per month on Sonnet, Opus, image, and video generation, this workstation moves recurring workloads to controlled, local infrastructure.

Paid off in approximately 16 months
Save approximately $47,124 over three years
Save approximately $104,931 over five years

Characteristics

GPU 2x RTX PRO 6000 Max-Q

CPU Intel Xeon 658X Enterprise (24 Cores)

Board ASUS Pro WS W890E-SAGE SE

VRAM 192 GB GDDR6 Max-Q

System RAM 256GB Kingston DDR5 ECC (64GB x 4)

Storage 2TB Boot NVMe + 2TB Active NVMe SSD

Mass Storage 12TB 7200RPM Enterprise HDD

Power 1700W Pro PSU

Cooling Air-cool GPUs · Liquid-cool CPU

SMTTR - Standard Matter by SMTTR - Standard Matter

SMTTR ONE 192GB VRAM NVIDIA Blackwell AI Workstation

SKU: SMTTR-WK5000-4 ✔ New, Factory-Sealed

Regular price: $35,075.00
Sale price: $35,075.00
Regular price: ~~$23,769.00~~

FREE Delivery on Orders $3,000+

In Stock: Only 2 Left

* NO SALES TAX Except in CA

** Prices and stock might change without notice due to high market volatility.

***Corporate & Lab Pricing Available - Get Up to 5% OFF on Volume

Run 120B MoE models at up to 130 tokens/s, generate SDXL images in 1.25 seconds, and produce 5-second 720p video clips locally. ... Read more

Run 120B MoE models at up to 130 tokens/s, generate SDXL images in 1.25 seconds, and produce 5-second 720p video clips locally. Compared with a $5,000 monthly AI bill, the workstation pays for itself in 8 months, saving $137,124 over three years and approximately $254,931 over five years.

VRAM:

192GB 96GB

GPUs (192GB total):

4x 5000 Blackwell 48GB 2x 6000 Blackwell Max-Q 96GB