Run LLMs in the cloud
Live GPU prices from Vast.ai and RunPod, refreshed every hour. Rent a card by the hour, or start from a model and see the cheapest GPU it fits on.
Which GPU should I rent?
16 GBruns up to ~14B models at Q4
NVIDIA RTX A400016 GB
$0.006 /GB·h
from $2.12/day
NVIDIA Tesla V100 PCIe 16GB16 GB
$0.006 /GB·h
from $2.45/day
NVIDIA T416 GB
$0.009 /GB·h
from $3.59/day
NVIDIA RTX 4000 Ada Generation20 GB
$0.010 /GB·h
from $4.80/day
24 GBruns up to ~32B models at Q4
NVIDIA RTX A500024 GB
$0.007 /GB·h
from $3.84/day
NVIDIA RTX PRO 4000 Blackwell24 GB
$0.011 /GB·h
from $6.45/day
NVIDIA L424 GB
$0.014 /GB·h
from $8.04/day
32 GBruns 32B models at Q4 with room for long context
NVIDIA RTX PRO 4500 Blackwell32 GB
$0.010 /GB·h
from $8.05/day
NVIDIA RTX 5000 Ada Generation32 GB
$0.012 /GB·h
from $9.16/day
$0.014 /GB·h
from $14/day
NVIDIA RTX PRO 4500 Blackwell Server Edition32 GB
$0.022 /GB·h
from $17/day
48 GBruns up to ~70B models at Q4
NVIDIA RTX A600048 GB
$0.007 /GB·h
from $7.92/day
NVIDIA A4048 GB
$0.007 /GB·h
from $8.40/day
NVIDIA RTX 6000 Ada Generation48 GB
$0.014 /GB·h
from $16/day
NVIDIA L4048 GB
$0.014 /GB·h
from $17/day
NVIDIA L40S48 GB
$0.016 /GB·h
from $19/day
80 GBruns 70B with long context, or ~120B MoE at Q4
NVIDIA A100 80GB SXM80 GB
$0.013 /GB·h
from $25/day
NVIDIA H100 PCIe80 GB
$0.025 /GB·h
from $48/day
NVIDIA H100 SXM80 GB
$0.034 /GB·h
from $65/day
96 GB+runs 120B MoE models, and 200B+ MoE on the largest cards
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition96 GB
$0.013 /GB·h
from $31/day
NVIDIA RTX PRO 6000 Blackwell Workstation Edition96 GB
$0.015 /GB·h
from $33/day
NVIDIA RTX PRO 6000 Blackwell Server Edition96 GB
$0.015 /GB·h
from $35/day
NVIDIA H200 SXM141 GB
$0.025 /GB·h
from $86/day
NVIDIA H200 NVL141 GB
$0.035 /GB·h
from $117/day
NVIDIA B200192 GB
$0.031 /GB·h
from $144/day
NVIDIA B300288 GB
$0.024 /GB·h
from $167/day
Daily prices assume 24 hours of nonstop use.
Prices refresh hourly. Each one is the median of the cheapest reliable hosts, not the single lowest listing.
Frequently Asked Questions
- What is the difference between Vast.ai Marketplace and RunPod Community Cloud?
Vast.ai Marketplace is individual hosts renting out their own machines. RunPod Community Cloud is RunPod's cheaper tier, still on hardware RunPod manages. Marketplace is usually the lowest price, but reliability depends on the host, so we only count verified hosts with a reliability score of 95% or higher. Community Cloud costs a little more and behaves more like a normal cloud.
- Where do the prices on this page come from?
Vast.ai prices are the median of the five cheapest offers from verified hosts with a reliability score of 95% or higher. RunPod prices are the lowest Community Cloud rate. Both refresh every hour. A median beats the single cheapest listing because rock-bottom offers often vanish within minutes. If a provider is unreachable, we keep its last known prices for up to 24 hours rather than dropping the GPU.
- Why is the price on RunPod's site higher than the one shown here?
RunPod's console opens on Secure Cloud, its pricier tier; the price here is for Community Cloud. Switch to Community Cloud on RunPod's deploy page and the number should match, give or take an hourly refresh. On Vast.ai the listing you land on can also differ, since individual hosts change prices and go offline through the day.
- Is it cheaper to rent a GPU or buy one?
Renting is cheaper below a few hundred hours of use. Buying wins once your total hours pass the card's price divided by the hourly rate; for a consumer card that is usually over a year of 8-hour days. Rent for short jobs, for an H100 you can't buy anyway, or to try a 70B model on a 48 GB card before buying one. Each GPU page shows its own breakeven.
- Do you earn a commission on the Deploy links?
Yes. Deploy buttons are partner links, and we may earn a commission at no extra cost to you. The commission does not affect the order: rows are sorted by price, and the cheapest provider for each GPU is always listed first, whether or not it pays us more.