Top Blackwell consumer GPU for local AI development and rendering.
Weekly median price per GPU per hour · Get the data
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 RTX 5090 has 32 GB of VRAM. In practice, that's enough memory for roughly 41B parameters at 4-bit or 9B 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) | RTX 5090s needed | Cost /hr | Cost /mo |
|---|---|---|---|---|
Qwen3.8-27B
27B
|
18 GB
|
1
|
$0.65
|
$468
|
Gemma 4 31B
30.7B
|
22 GB
|
1
|
$0.65
|
$468
|
GPT-OSS-120B
117B · 5.1B active
|
67 GB
|
3
|
$1.95
|
$1,404
|
GLM-5.3-Flash
320B · 18B active
|
178 GB
|
7
|
$4.55
|
$3,276
|
MiniMax-M3
428B · 23B active
|
239 GB
|
9
|
–
|
–
|
DeepSeek V4.1 Flash
552B · 16B active
|
306 GB
|
11
|
–
|
–
|
Kimi K3
2.8T · 104B active
|
1,544 GB
|
54
|
–
|
–
|
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 RTX 5090 · Per GPU
| Compute · dense | |
|---|---|
| FP4 | 1,676 TFLOPS 3,352 with sparsity |
| FP8 | 419 TFLOPS 838 with sparsity |
| FP16 / BF16 | 209.5 TFLOPS 419 with sparsity |
| INT8 | 838 TOPS 1,676 with sparsity |
| FP32 | 104.8 TFLOPS |
| Precision support | FP4FP8FP16BF16TF32FP32INT8 |
| Memory | |
|---|---|
| Capacity | 32 GB GDDR7 |
| Bandwidth | 1,792 GB/s |
| Bus width | 512-bit |
| ECC | No |
| Silicon | |
|---|---|
| Architecture | Blackwell |
| Process | TSMC 4N |
| Transistors | 92.2 billion |
| Shader cores | 21,760 CUDA cores |
| Matrix cores | 680 Tensor cores |
| Compute units | 170 SMs |
| Fabric and host | |
|---|---|
| Host interface | PCIe 5.0 |
| Power | |
|---|---|
| Board power | 575 W |
| Cooling | Active |
Source: official Nvidia RTX 5090 datasheet.
As of October 2, 2026, the median on-demand price is $0.65 per GPU per hour across 14 providers with a priced on-demand config.
| Billing type | Configs | Median /GPU/hr | Cheapest |
|---|---|---|---|
On-demand | 123 | $0.65 | $0.35 (in stock, HyperAI) |
Reserved | 93 | $0.61 | $0.21 (1 mo, in stock, HyperAI) |
Spot | 76 | $0.45 | $0.10 (in stock, Vast.ai) |
1 provider also quotes the RTX 5090 on a custom contract, priced per deal.
At 720 hours per month, one RTX 5090 costs an estimated $468 at the median on-demand price. Cheapest verified in stock: $252 per month on-demand, $151 reserved (1 mo), $72 spot.
GetDeploying currently tracks RTX 5090 configs from 22 providers. The cheapest verified in-stock on-demand configs come from HyperAI, GPU.ai, Vast.ai and Salad. See the full price comparison above for every provider and config.
One RTX 5090 runs models up to roughly 41B parameters at 4-bit quantization or 9B 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.
32GB GDDR7 with Blackwell FP4 Tensor Cores. Highest consumer GPU performance available. Strong for local AI development, LoRA fine-tuning, and image generation.
Consumer GeForce GPU. No ECC memory. Subject to Nvidia's consumer EULA, which may restrict data center use. For production workloads, consider the L40S or RTX 6000 Ada.
Nvidia RTX PRO 6000
96 GB GDDR7 · workstation card
Nvidia RTX 4090
24 GB GDDR6X · 44% less bandwidth than the RTX 5090
Nvidia L40S
48 GB GDDR6 · 52% less bandwidth than the RTX 5090
Last updated
