Vast.aiIn stock
On-Demand from $0.44
From
$0.44 / GPU / hr
On-Demand · California, USVisit website →
Rental price index
NVIDIA RTX 5090
🎮The consumer flagship. 32 GB and H100-class bandwidth at a fraction of the price.
Prices collected 2026-10-11 · 10 providers, 76 configurations
On-demandSpot
Index median price per GPU per hour, by billing type. Nothing is interpolated.
At a glance
Median on-demand
$0.81/GPU/hr
Cheapest on-demand
$0.44Vast.ai
Vast.aiFloor, any billing
$0.25incl. spot
Per month at 720h
$580
NVIDIA RTX 5090 pricing by provider
One card per provider, cheapest first. A configuration is one priced unit — an instance type, a SKU, a marketplace offer — and the per-GPU price is that unit's price divided by its GPU count. Open a card to see every configuration with its vCPU, RAM and region where the provider publishes them.
10 providers · 75 configurations
GPU.aiIn stock
AEUAE2 configs1xcollected 2026-10-11
On-Demand from $0.47
From
$0.47 / GPU / hr
On-DemandVisit website →
SaladCloudIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $0.50Spot from $0.25
From
$0.50 / GPU / hr
On-DemandVisit website →AkashIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $0.68
From
$0.68 / GPU / hr
On-DemandVisit website →
RunPod Community CloudIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $0.69
From
$0.69 / GPU / hr
On-DemandVisit website →
BeamIn stock
USUSA2 configs1xcollected 2026-10-11
On-Demand from $0.72
From
$0.72 / GPU / hr
On-DemandVisit website →SpheronIn stock
1 config1xcollected 2026-10-11
On-Demand from $0.78
From
$0.78 / GPU / hr
On-DemandVisit website →
RuncrateIn stock
1 config1xcollected 2026-10-11
On-Demand from $0.85
From
$0.85 / GPU / hr
On-DemandVisit website →
io.netIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $0.89
From
$0.89 / GPU / hr
On-DemandVisit website →
RunPod Secure CloudIn stock
USUSA1 config1xcollected 2026-10-11
On-Demand from $1.19
From
$1.19 / GPU / hr
On-DemandVisit website →Recent price changes
Every time a provider's median for the NVIDIA RTX 5090 moved between two consecutive snapshots. A listing appearing or vanishing is not a price change and is not shown.
| When | Provider | From | To | Change |
|---|---|---|---|---|
| 2026-10-11 09:14 | Vast.ai | $0.58 | $0.57 | -1.9% |
| 2026-10-10 08:53 | Vast.ai | $0.68 | $0.58 | -15.4% |
| 2026-10-10 08:53 | GPU.ai | $0.82 | $0.83 | +1.2% |
| 2026-10-09 09:31 | RunPod Secure Cloud | $0.99 | $1.19 | +20.2% |
| 2026-10-09 09:31 | Vast.ai | $0.60 | $0.68 | +13.2% |
| 2026-10-09 09:31 | GPU.ai | $0.74 | $0.82 | +10.9% |
| 2026-10-08 09:24 | Vast.ai | $0.57 | $0.60 | +4.8% |
| 2026-10-07 09:10 | Vast.ai | $0.60 | $0.57 | -4.6% |
| 2026-10-07 09:10 | GPU.ai | $0.71 | $0.73 | +2.8% |
| 2026-10-06 09:19 | GPU.ai | $0.73 | $0.71 | -2.1% |
Spot, reserved and where it is cheapest
Spot median
$0.2569% off · 1 providers
Reserved median
—mostly by quote
USA providers
$0.69median of 7
UAE providers
$0.83median of 1
What fits on a NVIDIA RTX 5090
Weights + KV cache at 8K context + runtime overhead, against 29 GB usable of the 32 GB on the card. Capacity follows total parameters, even for mixture-of-experts models — any expert may be needed next. Where one card is not enough, the table says how many are, and prices the set at today's median.
| Model | Params | 4-bit | 8-bit | 16-bit |
|---|---|---|---|---|
| 2.8T104B active | 55× GPU$44.28/hr | 104× GPU$83.72/hr | 195× GPU$156.98/hr | |
| 552B16B active | 11× GPU$8.86/hr | 21× GPU$16.91/hr | 39× GPU$31.40/hr | |
 Llama 4 Maverick | 400B17B active | 8× GPU$6.44/hr | 15× GPU$12.08/hr | 28× GPU$22.54/hr |
 gpt-oss-120b | 117B5.1B active | 3× GPU$2.42/hr | 5× GPU$4.03/hr | 9× GPU$7.25/hr |
| Llama 3.1 70B | 70B | 2× GPU$1.61/hr | 3× GPU$2.42/hr | 6× GPU$4.83/hr |
 Gemma 4 31B | 31B | ✓ 21.2 GB~82 tok/s · $2.73/M tok | 2× GPU$1.61/hr | 3× GPU$2.42/hr |
 Qwen3 Coder 30B-A3B | 30.5B3.3B active | ✓ 19 GB~772 tok/s · $0.29/M tok | 2× GPU$1.61/hr | 3× GPU$2.42/hr |
| Qwen3.8 27B | 27B | ✓ 18.7 GB~94 tok/s · $2.38/M tok | 2× GPU$1.61/hr | 2× GPU$1.61/hr |
| gpt-oss-20b | 21B3.6B active | ✓ 13.7 GB~708 tok/s · $0.32/M tok | ✓ 24.2 GB~375 tok/s · $0.6/M tok | 2× GPU$1.61/hr |
| Llama 3.1 8B | 8B | ✓ 6.9 GB~319 tok/s · $0.7/M tok | ✓ 10.9 GB~169 tok/s · $1.32/M tok | ✓ 18.4 GB~90 tok/s · $2.48/M tok |
tok/s is decode throughput from memory bandwidth, derated 20%, single stream — an estimate, not a benchmark. $/M tok is the hourly median divided by that rate: the one-user floor for a million generated tokens. Set cost uses today's median × GPUs required.
Buy or rent a RTX 5090?
Approx. list price
$1,999not tracked
Rent, per month
$580at median, 720 h
Break-even
2,483 h≈ 3.4 months
Rent for fewer than about 3.4 months of continuous use and renting wins on the card alone; past that, owning does — before the server, power, cooling and the person who keeps it running, which for a datacenter card are not small. The list price is a typical quoted figure and is not tracked by the index. Thinking of owning one? What the RTX 5090 runs locally · price in India.
Specifications
Memory32 GB GDDR7
Memory bandwidth1,792 GB/s
ArchitectureBlackwell
VendorNVIDIA
ClassConsumer
Released2025-01
Bandwidth is listed because it governs generation speed: each token streams the active weights once, so tokens per second is bandwidth divided by bytes read. TFLOPs decide prompt processing and training, not how fast text appears.
Alternatives and comparisons
RTX 5090 vs RTX 4090Price, memory, what fits, cost per tokenRTX 5090 vs RTX 5080Price, memory, what fits, cost per tokenRTX 5090 vs A100 80GBPrice, memory, what fits, cost per tokenRTX 5090 vs RTX PRO 6000 BlackwellPrice, memory, what fits, cost per token
NVIDIA RTX 409024 GB · $0.48/hr median · 13 providersNVIDIA RTX 309024 GB · $0.22/hr median · 7 providersNVIDIA RTX 408016 GB · $0.27/hr median · 6 providers
Questions
How much does it cost to rent a NVIDIA RTX 5090?
As of 2026-10-11, the median on-demand price across 10 providers is $0.81 per GPU-hour, about $580 a month at 720 hours. The cheapest we saw was $0.44 at Vast.ai.
Why do prices for the same GPU vary so much?
Hyperscalers bundle CPU, RAM and network into an instance price and charge for the ecosystem around it. Neoclouds sell the GPU more directly. Marketplaces are independent hosts competing on price, with the trade-offs of shared, third-party hardware. The per-GPU figure is the instance price divided by GPU count, which is the only way to compare an 8-GPU node with a single-card listing.
Is the median the price I will pay?
No — it is the middle of the market. Each provider's own median is taken first so a provider with many instance sizes counts once, then the median across providers. Spot and reserved rates are excluded from it; they set the floor shown separately. Use it to judge whether a quote is high or low, then verify with the provider.
Which models can a NVIDIA RTX 5090 run?
With 32 GB per GPU, the table above shows which reference models fit on one card at 4, 8 and 16-bit, using the same memory arithmetic as the rest of Runyard: weights plus KV cache plus runtime overhead, with capacity following total parameters even for mixture-of-experts models. Where a model needs more than one GPU it says how many and what that set costs at today's median.
Own the hardware instead? Work out what your card holds with the same arithmetic.
Check what fits →