NVIDIA H200 SXM

The Preferred Choice for Large Model Inference

Specifications for NVIDIA H200 SXM
ArchitectureHopper (H200)
VRAM141GB HBM3e
Mem Bandwidth4.8 TB/s
FP8 Tensor3,958 TFLOPS
Form FactorSXM5
TenancyDedicated
LocationMissoula, Montana

Ideal workloads: Inference on large foundation models (70B–405B parameters), RAG with large knowledge bases, long-context document processing, multi-modal AI, CubDen at scale.

Get H200 Pricing