Hardware & specs
Dedicated GPUs, routed across live inventory.
Rates below are our list rate card. Every launch is subject to live provider capacity. B200 is live-quoted and self-serve up to four GPUs when matching stock exists; larger datacenter clusters are arranged with our team.
NVIDIA B200
FP8 / FP4 Support
from $8.44
floor, quoted live per request
Memory
180GB HBM3e
Bandwidth
8.0 TB/s Bandwidth
Fabric
NVLink 5 · 1.8 TB/s
Node specification
- Architecture
- Blackwell
- Memory
- 180GB HBM3e
- Memory Bandwidth
- 8.0 TB/s
- Precision
- FP8 / FP4 / BF16
- Interconnect
- NVLink 5 · 1.8 TB/s
Specifications are NVIDIA chip specifications. The machine, interconnect and cooling depend on the supplier that serves the job.
| GPU | Memory | Bandwidth | Best for | $/GPU/hr |
|---|---|---|---|---|
| NVIDIA B200 | 180GB HBM3e | 8.0 TB/s Bandwidth | FP8 / FP4 Support | from $8.44 |
| NVIDIA H200 | 141GB HBM3e | 4.8 TB/s Bandwidth | Ultra-large LLM Inference | $5.20 |
| NVIDIA H100 SXM | 80GB HBM3 | 3.35 TB/s Bandwidth | Distributed Training Workhorse | $2.45 |
| NVIDIA A100 SXM | 80GB HBM2e | 2.0 TB/s Bandwidth | Cost-efficient Training | from $0.72 |
| NVIDIA L40S | 48GB GDDR6 | 864 GB/s Bandwidth | Fine-tuning & Vision | $1.10 |