NVIDIA B200 SXM
180 GB HBM3e · Blackwell · ~1000 W
Market Prices — on-demand $/GPU·hr
$5.49 – $14.00/GPU·hr (2.6× spread)
| Provider | GPU | $/GPU·hr | | Observed |
|---|
| RunPod | NVIDIA B200-SXM-180GB | $5.49 | | 1d ago |
| Lambda Labs | NVIDIA B200-SXM-180GB | $6.69 | | 1d ago |
| GCP (us-central1) | NVIDIA B200-SXM-180GB | $8.05 | | today |
| CoreWeave | NVIDIA B200-SXM-180GB | $8.60 | | 1d ago |
| OCI | NVIDIA B200-SXM-180GB | $14.00 | | today |
Efficiency by Workload
Offline scenario · on-demand pricing · best result per provider
llama2-70b
| Provider | $/GPU·hr | Perf/GPU | tok/$ | |
|---|
| RunPod | $5.49 | 12,357 tok/s | 8,103,115 tok/$ | |
| Lambda Labs | $6.69 | 12,357 tok/s | 6,649,641 tok/$ | |
| GCP (us-central1) | $8.05 | 12,357 tok/s | 5,522,793 tok/$ | |
| CoreWeave | $8.60 | 12,357 tok/s | 5,172,802 tok/$ | |
| OCI | $14.00 | 12,357 tok/s | 3,177,579 tok/$ | |
llama2-70b-interactive
| Provider | $/GPU·hr | Perf/GPU | tok/$ | |
|---|
| RunPod | $5.49 | 12,357 tok/s | 8,103,115 tok/$ | |
| Lambda Labs | $6.69 | 12,357 tok/s | 6,649,641 tok/$ | |
| GCP (us-central1) | $8.05 | 12,357 tok/s | 5,522,793 tok/$ | |
| CoreWeave | $8.60 | 12,357 tok/s | 5,172,802 tok/$ | |
| OCI | $14.00 | 12,357 tok/s | 3,177,579 tok/$ | |
mixtral-8x7b
| Provider | $/GPU·hr | Perf/GPU | tok/$ | |
|---|
| RunPod | $5.49 | 16,131 tok/s | 10,577,615 tok/$ | |
| Lambda Labs | $6.69 | 16,131 tok/s | 8,680,285 tok/$ | |
| GCP (us-central1) | $8.05 | 16,131 tok/s | 7,209,324 tok/$ | |
| CoreWeave | $8.60 | 16,131 tok/s | 6,752,454 tok/$ | |
| OCI | $14.00 | 16,131 tok/s | 4,147,936 tok/$ | |
stable-diffusion-xl
| Provider | $/GPU·hr | Perf/GPU | samp/$ | |
|---|
| RunPod | $5.49 | 4 samp/s | 2,490 samp/$ | |
| Lambda Labs | $6.69 | 4 samp/s | 2,043 samp/$ | |
| GCP (us-central1) | $8.05 | 4 samp/s | 1,697 samp/$ | |
| CoreWeave | $8.60 | 4 samp/s | 1,589 samp/$ | |
| OCI | $14.00 | 4 samp/s | 976 samp/$ | |