GPU model · price + stock · hourly snapshot
Nvidia H200 NVL cloud rental prices
3 providers list the Nvidia H200 NVL, 2 have it in stock. Prices run from $2.90 to $3.79 per GPU hour across 2 memory sizes.
LIVE Prices and stock checked 1 minute ago
Providers
Every provider renting the Nvidia H200 NVL
| Provider | From | Stock | Grab |
|---|---|---|---|
| Lium | $2.90 | In stock | Grab |
| Vast.ai | $3.27 | In stock | Grab |
| Runpod | $3.79 | Not reported | Grab |
In stock first, then cheapest. Stock is what the provider reported at our last check. One GPU for a month (730h) at $2.90: $2,117. Lium publishes live counts: 6 GPUs at $2.90, 67% rented.
The Nvidia H200 NVL sells out fast. Get one email when it is back in stock.
Price history
Price trend
$2.90 /GPU/hr · Oct 22
One point per day: the lowest listed $/GPU/hr across every provider we track, captured daily. 3 offers in the latest snapshot.
Configurations
All 3 configurations
| Configuration | VRAM | Per GPU/hr | Stock |
|---|---|---|---|
| 1x NVIDIA H200 NVLLium | — | $2.90 | In stock |
| 10x H200 NVLVast.ai | 1404 GB | $3.27 | In stock |
| 1x H200 NVLRunpod | 143 GB | $3.79 | Not reported |
Memory sizes
Nvidia H200 NVL price by memory size
The Nvidia H200 NVL comes in 2 memory sizes, priced apart. Check the size before you compare prices.
| Memory per GPU | Cheapest in stock | Cheapest listed | Providers |
|---|---|---|---|
| 140 GB | $3.27 Vast.ai | $3.27 | 1 |
| 143 GB | Out of stock | $3.79 | 1 |
Per GPU per hour, on-demand. Spot excluded.
Model fit
What fits in 141 GB
The question behind most rentals: will the model load. Sized against one Nvidia H200 NVL.
| Model | 16-bit | 8-bit | 4-bit |
|---|---|---|---|
| DeepSeek-V3 DeepSeek · 671B params · 37B active | 1,610 GB 12× · 82 GB free | 805 GB 6× · 41 GB free | 403 GB 3× · 20 GB free |
| Llama 3.1 405B Meta · 405B params | 972 GB 7× · 15 GB free | 486 GB 4× · 78 GB free | 243 GB 2× · 39 GB free |
| Qwen3 235B-A22B Alibaba · 235B params · 22B active | 564 GB 4× · 0 GB free | 282 GB 2× · 0 GB free | 141 GB 1× · 0 GB free |
| Mixtral 8x22B Mistral · 141B params · 39B active | 338 GB 3× · 85 GB free | 169 GB 2× · 113 GB free | 85 GB 1× · 56 GB free |
| Mistral Large 2 Mistral · 123B params | 295 GB 3× · 128 GB free | 148 GB 2× · 134 GB free | 74 GB 1× · 67 GB free |
| gpt-oss-120b OpenAI · 117B params · 5.1B active | 281 GB 2× · 1 GB free | 140 GB 1× · 1 GB free | 70 GB 1× · 71 GB free |
| Qwen2.5 72B Alibaba · 72.7B params | 174 GB 2× · 108 GB free | 87 GB 1× · 54 GB free | 44 GB 1× · 97 GB free |
| Llama 3.3 70B Meta · 70.6B params | 169 GB 2× · 113 GB free | 85 GB 1× · 56 GB free | 42 GB 1× · 99 GB free |
| Qwen3 32B Alibaba · 32.8B params | 79 GB 1× · 62 GB free | 39 GB 1× · 102 GB free | 20 GB 1× · 121 GB free |
| Gemma 2 27B Google · 27.2B params | 65 GB 1× · 76 GB free | 33 GB 1× · 108 GB free | 16 GB 1× · 125 GB free |
| Mistral Small 3 Mistral · 24B params | 58 GB 1× · 83 GB free | 29 GB 1× · 112 GB free | 14 GB 1× · 127 GB free |
| Llama 3.1 8B Meta · 8B params | 19 GB 1× · 122 GB free | 10 GB 1× · 131 GB free | 5 GB 1× · 136 GB free |
Required memory is weights plus 20% for KV cache and runtime at short context; long contexts and large batches need more. Cells tinted green fit on a single Nvidia H200 NVL.
Specifications
Nvidia H200 NVL specifications
| Architecture | Hopper |
|---|---|
| Memory per GPU | 141 GB HBM3e |
| Memory bandwidth | 4,800 GB/s |
| Release date | Q4 2024 |
| GPU interconnect | NVLink bridge 900 GB/s, 2 or 4 cards |
| FP8 compute rate | 1,671 TFLOPS (dense), 3,341 with sparsity |
| FP16 / BF16 compute rate | 835.5 TFLOPS (dense), 1,671 with sparsity |
| INT8 compute rate | 1,671 TOPS (dense), 3,341 with sparsity |
| Process node | TSMC 4N |
| Board power | 600 W |
| Partitioning | MIG, up to 7 instances |
Alternatives