NVIDIA · Hopper
H200 SXM 141 GB rental prices and specifications
A Hopper datacenter accelerator with 141 GB of HBM3e and 4.8 TB/s memory bandwidth, designed for memory-intensive training and large-model inference.
Profile checked · Primary source ↗
- Memory
- 141 GB
- Bandwidth
- 4800 GB/s
- Current offers
- 0
- FP16 dense
- 989.5 TFLOPS
- BF16 dense
- 989.5 TFLOPS
- FP8 dense
- 1979 TFLOPS
- Interconnect
- NVLink up to 900 GB/s; PCIe Gen5 128 GB/s.
Compare H200 SXM 141 GB cloud rentals
Use the exact 141 GB GPU profile when comparing rental quotes. Tables separate currencies, billing periods and rental terms; the required GPU count determines the whole-instance compute cost. Storage, transfer, taxes and other charges can be additional.
Source checks show when each price was actually observed. Historical quotes remain available separately and do not enter current price comparisons. Explore the sourced GPU rental price index or check a rental against your model requirements.
Dense tensor peaks are derived as one half of the manufacturer table values explicitly marked with sparsity. Vendor rounding is retained; these are theoretical peaks, not inference measurements. · Specification source 1 ↗
Recently checked listings
Current rental offers
| GPU / provider | Instance price / offer | Terms / location |
|---|---|---|
| No offers match these filters. Try another billing period or currency. | ||
Previously observed offers
21 retained quotes with an old check, a different allocation, or unavailable status. These are excluded from current prices and capacity counts.
Show previously observed offers →Measured performance
Inference benchmarks
| Model | GPU setup | Engine | Quant. | Workload | Reported tok/s | Evidence | Max-load capacity scenarioUSD / million output tokens |
|---|---|---|---|---|---|---|---|
| Qwen3-32B | H200 SXM 141 GB8 GPUs | vLLM with NVIDIA DynamoMeasured build undisclosed; pinned guide example uses Dynamo vllm-runtime:1.5.0 and optimizer target vLLM 0.12.0 | FP8Qwen/Qwen3-32B-FP8 | 4000 → 500input → output | 3462.0 | Source ↗Published Sep 21, 2026 | —No recent non-reference on-demand USD offer matches this GPU variant and count.8 H200 SXM GPUs (walkthrough system h200_sxm), four TP2 workers behind one frontend |
Output throughput is aggregate maximum-load performance, not single-user generation speed. Workloads show input → output tokens. Compare rows only when the model, GPU count, quantization, engine, and workload match.
Capacity scenarios combine an exact GPU variant/count rental with published output throughput from a different system. Throughput was not measured on the selected rental. Taxes, extras, idle time and minimum billing are excluded; these are not measured operating costs. Calculation guide.