NVIDIA H200
141 GB HBM3e · Hopper · SXM / NVL
Market Prices — on-demand $/GPU·hr
$4.29 – $10.60/GPU·hr (2.5× spread)
| Provider | GPU | $/GPU·hr | | Observed |
|---|
| Crusoe | NVIDIA H200-SXM-141GB | $4.29 | | 1d ago |
| RunPod | NVIDIA H200-SXM-141GB | $4.31 | | 1d ago |
| CoreWeave | NVIDIA H200-SXM-141GB | $6.31 | | 1d ago |
| OCI | NVIDIA H200-SXM-141GB | $10.00 | | today |
| AWS (us-east-1) | NVIDIA H200-NVL-141GB | $10.60 | | today |
| Azure (eastus2) | NVIDIA H200-SXM-141GB | $10.60 | | today |
| GCP (us-central1) | NVIDIA H200-SXM-141GB | $10.60 | | today |
Efficiency by Workload
Offline scenario · on-demand pricing · best result per provider
gptj
| Provider | $/GPU·hr | Perf/GPU | tok/$ | |
|---|
| Crusoe | $4.29 | 2,667 tok/s | 2,237,895 tok/$ | |
| RunPod | $4.31 | 2,667 tok/s | 2,227,510 tok/$ | |
| CoreWeave | $6.31 | 2,667 tok/s | 1,521,485 tok/$ | |
| OCI | $10.00 | 2,667 tok/s | 960,057 tok/$ | |
| Azure (eastus2) | $10.60 | 2,667 tok/s | 905,714 tok/$ | |
| GCP (us-central1) | $10.60 | 2,667 tok/s | 905,640 tok/$ | |
llama2-70b
| Provider | $/GPU·hr | Perf/GPU | tok/$ | |
|---|
| Crusoe | $4.29 | 4,432 tok/s | 3,718,846 tok/$ | |
| RunPod | $4.31 | 4,432 tok/s | 3,701,589 tok/$ | |
| CoreWeave | $6.31 | 4,432 tok/s | 2,528,344 tok/$ | |
| OCI | $10.00 | 4,432 tok/s | 1,595,385 tok/$ | |
| Azure (eastus2) | $10.60 | 4,432 tok/s | 1,505,080 tok/$ | |
| GCP (us-central1) | $10.60 | 4,432 tok/s | 1,504,958 tok/$ | |
| AWS (us-east-1) | $10.60 | 3,894 tok/s | 1,322,401 tok/$ | |
llama2-70b-interactive
| Provider | $/GPU·hr | Perf/GPU | tok/$ | |
|---|
| Crusoe | $4.29 | 4,437 tok/s | 3,723,619 tok/$ | |
| RunPod | $4.31 | 4,437 tok/s | 3,706,340 tok/$ | |
| CoreWeave | $6.31 | 4,437 tok/s | 2,531,589 tok/$ | |
| OCI | $10.00 | 4,437 tok/s | 1,597,432 tok/$ | |
| Azure (eastus2) | $10.60 | 4,437 tok/s | 1,507,012 tok/$ | |
| GCP (us-central1) | $10.60 | 4,437 tok/s | 1,506,889 tok/$ | |
| AWS (us-east-1) | $10.60 | 3,917 tok/s | 1,330,251 tok/$ | |
mixtral-8x7b
| Provider | $/GPU·hr | Perf/GPU | tok/$ | |
|---|
| Crusoe | $4.29 | 7,939 tok/s | 6,662,381 tok/$ | |
| RunPod | $4.31 | 7,939 tok/s | 6,631,465 tok/$ | |
| CoreWeave | $6.31 | 7,939 tok/s | 4,529,574 tok/$ | |
| OCI | $10.00 | 7,939 tok/s | 2,858,161 tok/$ | |
| Azure (eastus2) | $10.60 | 7,939 tok/s | 2,696,379 tok/$ | |
| GCP (us-central1) | $10.60 | 7,939 tok/s | 2,696,159 tok/$ | |
resnet50
| Provider | $/GPU·hr | Perf/GPU | samp/$ | |
|---|
| Crusoe | $4.29 | 96,662 samp/s | 81,115,385 samp/$ | |
| RunPod | $4.31 | 96,662 samp/s | 80,738,979 samp/$ | |
| CoreWeave | $6.31 | 96,662 samp/s | 55,148,177 samp/$ | |
| OCI | $10.00 | 96,662 samp/s | 34,798,500 samp/$ | |
| Azure (eastus2) | $10.60 | 96,662 samp/s | 32,828,774 samp/$ | |
| GCP (us-central1) | $10.60 | 96,662 samp/s | 32,826,099 samp/$ | |
retinanet
| Provider | $/GPU·hr | Perf/GPU | samp/$ | |
|---|
| Crusoe | $4.29 | 1,900 samp/s | 1,594,437 samp/$ | |
| RunPod | $4.31 | 1,900 samp/s | 1,587,038 samp/$ | |
| CoreWeave | $6.31 | 1,900 samp/s | 1,084,015 samp/$ | |
| OCI | $10.00 | 1,900 samp/s | 684,013 samp/$ | |
| Azure (eastus2) | $10.60 | 1,900 samp/s | 645,296 samp/$ | |
| GCP (us-central1) | $10.60 | 1,900 samp/s | 645,243 samp/$ | |
| AWS (us-east-1) | $10.60 | 1,725 samp/s | 585,972 samp/$ | |
stable-diffusion-xl
| Provider | $/GPU·hr | Perf/GPU | samp/$ | |
|---|
| Crusoe | $4.29 | 2 samp/s | 2,040 samp/$ | |
| RunPod | $4.31 | 2 samp/s | 2,030 samp/$ | |
| CoreWeave | $6.31 | 2 samp/s | 1,387 samp/$ | |
| OCI | $10.00 | 2 samp/s | 875 samp/$ | |
| Azure (eastus2) | $10.60 | 2 samp/s | 825 samp/$ | |
| GCP (us-central1) | $10.60 | 2 samp/s | 825 samp/$ | |
| AWS (us-east-1) | $10.60 | 2 samp/s | 712 samp/$ | |
bert
| Provider | $/GPU·hr | Perf/GPU | samp/$ | |
|---|
| Crusoe | $4.29 | 9,224 samp/s | 7,740,315 samp/$ | |
| RunPod | $4.31 | 9,224 samp/s | 7,704,397 samp/$ | |
| CoreWeave | $6.31 | 9,224 samp/s | 5,262,433 samp/$ | |
| OCI | $10.00 | 9,224 samp/s | 3,320,595 samp/$ | |
| Azure (eastus2) | $10.60 | 9,224 samp/s | 3,132,637 samp/$ | |
| GCP (us-central1) | $10.60 | 9,224 samp/s | 3,132,382 samp/$ | |
dlrm-v2
| Provider | $/GPU·hr | Perf/GPU | samp/$ | |
|---|
| Crusoe | $4.29 | 80,470 samp/s | 67,526,853 samp/$ | |
| RunPod | $4.31 | 80,470 samp/s | 67,213,503 samp/$ | |
| CoreWeave | $6.31 | 80,470 samp/s | 45,909,699 samp/$ | |
| OCI | $10.00 | 80,470 samp/s | 28,969,020 samp/$ | |
| Azure (eastus2) | $10.60 | 80,470 samp/s | 27,329,264 samp/$ | |
| GCP (us-central1) | $10.60 | 80,470 samp/s | 27,327,038 samp/$ | |
3d-unet
| Provider | $/GPU·hr | Perf/GPU | samp/$ | |
|---|
| Crusoe | $4.29 | 7 samp/s | 5,739 samp/$ | |
| RunPod | $4.31 | 7 samp/s | 5,713 samp/$ | |
| CoreWeave | $6.31 | 7 samp/s | 3,902 samp/$ | |
| OCI | $10.00 | 7 samp/s | 2,462 samp/$ | |
| Azure (eastus2) | $10.60 | 7 samp/s | 2,323 samp/$ | |
| GCP (us-central1) | $10.60 | 7 samp/s | 2,323 samp/$ | |
| AWS (us-east-1) | $10.60 | 6 samp/s | 2,097 samp/$ | |