Specs, live pricing, and best-fit workloads, side by side.
| Spec | H100 80GB | A100 40GB |
|---|---|---|
| VRAM | 80GB | 40GB |
| Tier | Fast Serving | Training GPU |
| Live from | $2.49/hr | $1.25/hr |
| Fits | 32B+ inference, high-throughput 8B/14B, production endpoints | 7B training, LoRA multi-card, medium-scale workloads |
| Best for | Production AI endpoints, large model serving, high-performance computing | Model training, fine-tuning, LoRA adaptation |
| Can serve models | Yes | Yes |
Live-from prices are last-seen estimates, rechecked before launch.