Live Floor
$3.8988
Vast · US
The NVIDIA B200 is built on the Blackwell architecture with 192 GB of HBM3e VRAM.
Best live route via Vast · US
Live Floor
$3.8988
Vast · US
Average Price
$13.3522
Across 8 live offers
Reliability
98.2%
Based on 8 scored hosts
Last Day Delta
+$0.0000
Latest daily low: 2026-05-20
30-Day Price Trend
Available Offers
Sorted by cheapest hourly rate
| Provider | Price/hr | VRAM | GPUs | Reliability | Region | |
|---|---|---|---|---|---|---|
| Vast | $3.8988 | 179.1 GB | 1x | 99.6% | US | View Deal |
| Vast | $4.0405 | 179.1 GB | 1x | 99.1% | US | View Deal |
| RunPod | $5.9800 | 180 GB | 1x | 90.0% | -- | View Deal |
| Vast | $7.7963 | 179.1 GB | 2x | 99.6% | US | View Deal |
| Vast | $15.5911 | 179.1 GB | 4x | 99.6% | US | View Deal |
| Vast | $15.6424 | 179.1 GB | 4x | 99.4% | US | View Deal |
| Vast | $17.9598 | 179.1 GB | 2x | 99.0% | US | View Deal |
| Vast | $35.9085 | 179.1 GB | 4x | 99.0% | US | View Deal |
GPU Overview
The NVIDIA B200 is built on the Blackwell architecture with 192 GB of HBM3e VRAM. Designed for datacenter-scale deployment, it delivers industry-leading memory capacity and compute throughput for scalable AI infrastructure and distributed training clusters.
With 192 GB of VRAM, the B200 handles Large Language Models (LLM) and fine-tuning of the largest foundation models without model parallelism constraints. Its benchmark score of 120 reflects uncompromising performance for enterprise-grade AI, making it a top-tier choice for organizations running production inference and large-scale training pipelines.