Live market index
GPU prices on Vast.ai
Updated · 45 priced GPUs · refreshed hourly
The cheapest GPU on Vast.ai right now is RTX 3060 (12 GB) at $0.04/hr — about $29/month if it ran 24/7. Its median price is $0.08/hr ($58/month). Every price below is a live whole-card offer on the Vast.ai marketplace, refreshed hourly; a card is listed once at least five whole-card offers exist at the same time.
All rentable GPUs, cheapest first
| GPU | VRAM | Live offers | From $/hr | Median $/hr | From $/month | Median $/month |
|---|---|---|---|---|---|---|
| RTX 3060 | 12 GB | 15 | $0.04 | $0.08 | $29 | $58 |
| GTX 1660 Ti | 6 GB | 4 | $0.04 | $0.07 | $29 | $51 |
| GTX 1660 S | 6 GB | 3 | $0.05 | $0.05 | $37 | $37 |
| RTX 3060 Ti | 8 GB | 8 | $0.06 | $0.11 | $44 | $80 |
| GTX 1080 Ti | 11 GB | 1 | $0.07 | $0.07 | $51 | $51 |
| GTX 1080 | 8 GB | 1 | $0.08 | $0.08 | $58 | $58 |
| RTX A4000 | 16 GB | 4 | $0.09 | $0.10 | $66 | $73 |
| RTX 2080 Ti | 11 GB | 5 | $0.09 | $0.10 | $66 | $73 |
| RTX 5060 | 8 GB | 2 | $0.09 | $0.09 | $66 | $66 |
| RTX 3090 | 24 GB | 24 | $0.10 | $0.24 | $73 | $175 |
| L4 | 24 GB | 7 | $0.10 | $0.10 | $73 | $73 |
| RTX 5060 Ti | 16 GB | 10 | $0.10 | $0.16 | $73 | $117 |
| RTX 5060 Ti | 8 GB | 6 | $0.10 | $0.11 | $73 | $80 |
| RTX 3070 | 8 GB | 3 | $0.10 | $0.18 | $73 | $131 |
| RTX 4070 | 12 GB | 4 | $0.11 | $0.15 | $80 | $110 |
| RTX 3080 | 10 GB | 20 | $0.11 | $0.14 | $80 | $102 |
| RTX 3080 Ti | 12 GB | 11 | $0.12 | $0.19 | $88 | $139 |
| RTX 4060 Ti | 8 GB | 4 | $0.12 | $0.15 | $88 | $110 |
| RTX 5070 Ti | 16 GB | 16 | $0.13 | $0.24 | $95 | $175 |
| RTX 5070 | 12 GB | 8 | $0.13 | $0.16 | $95 | $117 |
| RTX 4090 | 24 GB | 24 | $0.14 | $0.42 | $102 | $307 |
| RTX 4070 Ti | 12 GB | 2 | $0.15 | $0.19 | $110 | $139 |
| RTX 4060 Ti | 16 GB | 4 | $0.16 | $0.17 | $117 | $124 |
| RTX A2000 | 6 GB | 3 | $0.16 | $0.27 | $117 | $197 |
| RTX 4000Ada | 20 GB | 2 | $0.18 | $0.19 | $131 | $139 |
| RTX 4080 | 16 GB | 11 | $0.18 | $0.26 | $131 | $190 |
| RTX 4070S Ti | 16 GB | 3 | $0.19 | $0.22 | $139 | $161 |
| RTX 3070 Ti | 8 GB | 1 | $0.19 | $0.19 | $139 | $139 |
| RTX 3090 Ti | 24 GB | 4 | $0.22 | $0.27 | $161 | $197 |
| RTX 4080S | 16 GB | 4 | $0.22 | $0.24 | $161 | $175 |
| RTX 5080 | 16 GB | 15 | $0.23 | $0.31 | $168 | $226 |
| CMP 170HX | 64 GB | 1 | $0.24 | $0.24 | $175 | $175 |
| RTX PRO 4000 | 24 GB | 4 | $0.26 | $0.26 | $190 | $190 |
| RTX PRO 5000 | 48 GB | 4 | $0.35 | $0.72 | $255 | $526 |
| RTX A6000 | 48 GB | 4 | $0.35 | $0.44 | $255 | $321 |
| RTX PRO 4500 | 32 GB | 5 | $0.35 | $0.38 | $255 | $277 |
| RTX 5090 | 32 GB | 30 | $0.37 | $0.60 | $270 | $438 |
| A100 SXM4 | 40 GB | 4 | $0.48 | $0.80 | $350 | $584 |
| RTX 4090 | 48 GB | 0 | $0.68 | $1.34 | $496 | $978 |
| RTX 6000 Ada | 48 GB | 1 | $0.74 | $0.74 | $540 | $540 |
| A100 PCIE | 40 GB | 1 | $0.74 | $0.74 | $540 | $540 |
| RTX PRO 6000 WS | 96 GB | 12 | $0.87 | $1.42 | $635 | $1,037 |
| RTX PRO 6000 Max-Q | 96 GB | 7 | $1.14 | $1.34 | $832 | $978 |
| RTX PRO 6000 S | 96 GB | 2 | $1.60 | $1.60 | $1,168 | $1,168 |
| H200 | 141 GB | 1 | $2.63 | $2.63 | $1,920 | $1,920 |
How this index works
- Source: the Vast.ai marketplace, read through the same catalog that powers CUDAnetes deployments.
- Refresh: hourly. Each card keeps its last known price until a newer cycle succeeds, so a stale row is possible right after an outage.
- Whole cards only: shared or fractional GPU offers are excluded, and a card is listed once at least five simultaneous whole-card offers exist.
- Datacenter cards: several high-end GPUs are sold on Vast.ai mostly as multi-GPU nodes or fractional slices. Only offers for one whole card count here, so a card like H100 SXM can be absent while the hardware is rentable.
- Two numbers per card: the lowest live offer and the median of all live offers — the median is what a typical rental costs.
- Monthly cost: hourly rate × 730, i.e. the same GPU running 24/7 for an average month.
Comparing providers? See Vast.ai vs RunPod, side by side.
Ready to rent one? Step-by-step guides: vLLM, Ollama, ComfyUI — or all deploy guides.
Dated snapshot: November 2026 · machine-readable: /gpu-prices.json
Questions
What is the cheapest GPU on Vast.ai right now?
RTX 3060 with 12 GB is the cheapest rentable card in the current snapshot: $0.04 per hour, about $29 per month if it ran 24/7. That is the lowest live offer; the median for this card is $0.08 per hour ($58 per month).
What is the difference between the lowest and the median price?
The lowest price is the cheapest single live offer for that GPU. The median is the middle price across all live offers of the same card, so it is closer to what you will actually pay: cheap hosts are fewer and tend to be less reliable, physically further away, or slower to start.
How often does this index update?
Market data is refreshed every hour from the Vast.ai marketplace, and the timestamp at the top shows the last refresh. Only GPUs with at least five simultaneous whole-card offers are listed, so a card can drop off this page when supply dries up and return later.
Are these the prices I will be charged?
These are Vast.ai's own hourly rates, billed by Vast.ai to your own Vast account per second. CUDAnetes charges a flat orchestration subscription on top (free tier or $19 per month) and takes no markup on GPU time.
Why is the monthly figure based on 730 hours?
730 hours is an average month (365 days ÷ 12 × 24 hours), which makes hourly rates comparable to monthly cloud pricing. Your real bill depends on how long instances actually run.
Rent at these prices, serve behind one URL
CUDAnetes deploys any Docker container on these GPUs, keeps it healthy, scales it with load and gives you one public HTTPS endpoint. Free plan: 1 deployment, 5 GPUs, no card. You pay Vast.ai's hourly rate directly — nothing is added on top of GPU time.
Start free →