Rack-scale system for training and serving large AI models.
By provider shows one card per company, placed where its best offer ranked. By configuration lists every offer.
What you can rent comes first: in stock, then waitlist, then not reported, then out of stock. Priced offers rank ahead of quote-only, and on-demand ahead of other billing types.
Within a group, five factors set the order:
We match the provider, its country, GPU model, form factor, billing type, availability and instance name. Matching is partial and case-insensitive. Everyday words work too, so "interruptible" finds spot and "sold out" finds out of stock.
No providers match .
Heads up: A provider's own page may quote a different figure, for example on tax, region or a promotion. A monthly-only plan shows a derived hourly rate. Verify before provisioning. More on how we price.
One Vera Rubin has 288 GB of VRAM. In practice, that's enough memory for roughly 460B parameters at 4-bit or 125B at 16-bit, assuming a 32K context. Below are some open-weight LLMs, with the estimated memory and GPUs each one needs.
| Model | Memory (INT4 / FP4) | Vera Rubins needed | Cost /hr | Cost /mo |
|---|---|---|---|---|
Qwen3.8-27B
27B
|
18 GB
|
1
|
–
|
–
|
Gemma 4 31B
30.7B
|
22 GB
|
1
|
–
|
–
|
GPT-OSS-120B
117B · 5.1B active
|
67 GB
|
1
|
–
|
–
|
GLM-5.3-Flash
320B · 18B active
|
178 GB
|
1
|
–
|
–
|
MiniMax-M3
428B · 23B active
|
239 GB
|
1
|
–
|
–
|
DeepSeek V4.1 Flash
552B · 16B active
|
306 GB
|
2
|
–
|
–
|
Kimi K3
2.8T · 104B active
|
1,544 GB
|
6
|
–
|
–
|
Estimates based on the median on-demand rate. Memory is weights plus FP8 KV cache at 32K context per request (FP16/BF16 in the 16-bit column). GPU counts assume 90% of advertised VRAM is usable. No guarantee of runtime support, usable performance, or that a matching quantized build exists. Pricing methodology.
Nvidia Vera Rubin · Per GPU
| Compute · dense | |
|---|---|
| FP4 | 35,000 TFLOPS |
| FP8 | 17,500 TFLOPS |
| FP16 / BF16 | 4,000 TFLOPS |
| INT8 | 250 TOPS |
| FP32 | 130 TFLOPS |
| FP64 | 33 TFLOPS |
| Precision support | FP4FP6FP8FP16BF16TF32FP32FP64INT8 |
| Memory | |
|---|---|
| Capacity | 288 GB HBM4 |
| Bandwidth | 19,200 GB/s |
| Silicon | |
|---|---|
| Architecture | Rubin |
| Transistors | 336 billion |
| Compute units | 224 SMs |
| Fabric and host | |
|---|---|
| GPU interconnect | NVLink 3,000 GB/s |
| Host interface | PCIe 6.0 x16 |
| Power | |
|---|---|
| Cooling | Liquid |
Source: official Nvidia Vera Rubin datasheet.
As of November 9, 2026, we track 7 configs from 6 providers, none of them on demand. Prices are per GPU per hour.
| Billing type | Configs | Cheapest |
|---|---|---|
Reserved | 2 | $11.05 (60 mo, pre-order, Impossible Cloud) |
No median for reserved: we only show this when at least 3 providers list the GPU on that billing type. 5 providers also quote the Vera Rubin on a custom contract, priced per deal.
GetDeploying currently tracks Vera Rubin configs from 6 providers. See the full price comparison above for every provider and config.
One Vera Rubin runs models up to roughly 460B parameters at 4-bit quantization or 125B at 16-bit, assuming a 32K context. Larger models run across multiple GPUs: the model table above shows the estimated memory and GPU count for popular open-weight LLMs.
288GB of HBM4 per GPU in a 72-GPU NVLink domain. Suited to frontier training and long-context reasoning inference where memory bandwidth and NVLink throughput are the bottleneck.
A model that fits on one 8-GPU server does not need a 72-GPU NVLink domain, and the B300 has the same 288GB per GPU in standard x86 servers. For rack-scale work not limited by memory or NVLink bandwidth, the GB300 is the same 72-GPU design one generation back. Vera Rubin is for training and serving jobs too large for a single node.
Nvidia B300
288 GB HBM3e · 58% less bandwidth than the Vera Rubin
Nvidia GB300
288 GB HBM3e · 58% less bandwidth than the Vera Rubin
Nvidia GB200
186 GB HBM3e · 58% less bandwidth than the Vera Rubin
Last updated
