AI Compute RadarAll GPU classes · Price comparison
Vast.ai vs RunPod
What an hour of each GPU class costs on the two places people compare most, read from their public listings every hour: the Vast.ai median against RunPod's lowest posted price right now, and over the last thirty complete days a count of the days each was cheaper. Prices only; nothing here says which service is better.
The last 30 complete days
For every class, each complete day on which both sides have a usable daily value (Sep 21 to Oct 6): the median of those days on each side, and the number of days on which each side was the cheaper one. Daily values are the median of the day's hourly readings, glitch days excluded, the same series the Rent Index and the alerts use. The partial current day is left out.
| GPU | Vast.ai, median of days | RunPod, median of days | RunPod vs. Vast.ai | Days cheaper |
|---|---|---|---|---|
| RTX 3090 | collecting — 1 of 4 shared days | |||
| RTX 4090 | $0.48 | $0.34 | 29 % below | RunPod 16 · Vast.ai 016 days compared |
| RTX 5090 | $0.61 | $0.69 | 13 % above | RunPod 2 · Vast.ai 1416 days compared |
| L40S | collecting — 1 of 4 shared days | |||
| A100 | $0.80 | $1.00 | 25 % above | RunPod 0 · Vast.ai 1616 days compared |
| H100 | $2.75 | $2.59 | 6 % below | RunPod 11 · Vast.ai 216 days compared · 3 within a cent |
| H200 | $4.80 | $3.59 | 25 % below | RunPod 16 · Vast.ai 016 days compared |
| B200 | $8.13 | $5.98 | 26 % below | RunPod 16 · Vast.ai 016 days compared |
A day counts for RunPod when its lowest price sat below the Vast.ai median of that day, and for Vast.ai when the median sat below RunPod's price. Only days with a value on both sides are compared, so the day counts differ per class; RunPod has been read since 21 September 2026, the RTX 3090 and L40S since 6 October.
Two kinds of numbers
Vast.ai is a marketplace: hosts list their own machines at their own prices. The number here is the median over the verified offers a renter could take at the top of the hour, which is why an offer count stands next to it. Half the offers are cheaper than the median and half dearer; the cheapest one is shown as well.
RunPod posts one price per GPU type. The number here is the lowest on-demand price among the types that fold into a class, across its secure and community cloud. It is a floor, not a middle: the cheapest RunPod option, named by type.
So the comparison sets a floor against a middle, and says so. When RunPod's lowest price is above the Vast.ai median, RunPod is the dearer place for that class by any reading of the data. When it is below, the cheapest RunPod option undercuts the typical Vast.ai offer, while the cheapest Vast.ai offer may still be lower.
What this page does not say
- – Whether a machine is free at the moment you want it: Vast.ai's offer count is a measure of supply, RunPod's stock status is not read into these numbers.
- – How fast, how reliable or how well supported either side is. Nothing here was benchmarked.
- – What you pay in the end: storage, network egress, idle time and the pricing of interruptible (spot) capacity are outside these readings.
- – Which exact card you get inside a class on Vast.ai: the median spans PCIe, SXM and NVL variants where a class has them; RunPod's type is named.
- – That tomorrow looks like today. Every number is a reading with its time; the site describes and never forecasts.