NVIDIA B200
Data center GPU · Blackwell
Widely deployed, powerful, with lots of VRAM. The B200 is a great fit for large-scale training or inference workloads. While its FP4 performance is exceeded by the newer B300 GPU, the B200 wins when it comes to FP64 and INT8 operations.
The B200 is also available in its GB200 "Grace Blackwell" configuration, pairing B200 GPUs with NVIDIA's Grace CPU.
Tech specs
Cloud rental prices
14 providers · available in 8 countries · lowest price per GPU, per hour
| Provider | On-demand | Reserved | Spot |
|---|---|---|---|
 AWS | $14.24 | — | — |
 Beyond.pl | — | $5.10 | — |
 Cirrascale | — | $4.79 | — |
 CoreWeave | $8.60 | — | $4.26 |
| | — | $7.09 | $4.26 |
 Hyperstack | $6.00 | $5.10 | — |
 Lambda | $6.69 | — | — |
 Nebius | $7.15 | — | $3.95 |
 Oracle | $14.00 | — | — |
 Runpod | $5.89 | — | — |
 Together AI | $8.19 | — | — |
 Vast.ai | $6.56 | — | $6.25 |
 Verda | $6.11 | $5.62 | $2.14 |
 Vultr | $3.50 | — | — |
Models that fit in VRAM
Open-weights models that fit in VRAM. Estimated using 🤗 accelerate, plus approximation for up to 8K context.
Media
Peak theoretical performance
Performance figures assume no sparsity; in cases where only sparse performance figures are published by the manufacturer, these are halved to give approximate dense performance.
Resources
Detailed documentation from the manufacturer.
Similar GPUs
Other accelerators you might compare.