In our read of October 19, 2026 at 05:39 UTC, RunPod and Vast.ai quoted 24 of the same GPUs. On the 23 of them with a live quote on both sides of the community market, Vast.ai’s marketplace floor was the lower figure on 15 and RunPod’s Community Cloud rate on 8. On the RTX 4090, RunPod’s Community Cloud read $0.34 a GPU-hour and Vast.ai’s floor $0.4408.
Two different markets, compared apart
The question “which is cheaper” has a catch: on none of the 24 shared GPUs do both list an on-demand rate card. RunPod prints two fixed prices per card, Secure Cloud (its on-demand rate card) and Community Cloud (machines third parties host). Vast.ai prints no list price at all; it is a marketplace, and we record its floor, the median of the five cheapest verified offers at the minute of reading. A floor moves with supply, and an offer on it can be taken back. So our comparison page gives no on-demand verdict for this pair and compares Community Cloud with the Vast.ai floor on their own.
Card by card, October 19
RunPod’s Community Cloud was the lower figure on the RTX 4090 ($0.34 against $0.4408, a 4-GPU machine on Vast.ai), the B200 ($5.98 against $9.0026, a 4-GPU machine), the H200 ($3.59 against $5.0011) and, by about a cent, the L40S ($0.79 against $0.8008). Vast.ai’s floor was the lower one on the A100 80GB: $1.0022 against $1.19. On the RTX 3080 Ti the two round to the same $0.18 on the page; the floor read $0.1785, so it counts as Vast.ai’s.
The difference on the B200 is $9.0026 - $5.98 = $3.02 a GPU-hour; on the A100 80GB it is $1.19 - $1.0022 = $0.19. On all 15 cards where Vast.ai read lower, its floor read no higher than $1.2378 a GPU-hour, the RTX PRO 6000 Blackwell on a 2-GPU machine; all 23 rows are on the comparison page with both figures and their links.
The H100 SXM is not in the count. RunPod listed it at $2.69 on Community Cloud and $3.99 on Secure Cloud, while Vast.ai’s H100 SXM floor was withheld at that read as a thin market: too few verified offers to call a floor. A withheld floor enters no comparison.
Against the rest of the market
Both RunPod columns can be set beside the median of on-demand rate cards our card pages print, which leaves out every floor and community rate. The RTX 4090 median was $0.48 over 16 rate cards; RunPod’s Secure Cloud $0.89 sits above it, its Community Cloud $0.34 below. The A100 80GB median was $2.00 over 58 rate cards; RunPod’s Secure Cloud read $1.79 and its Community Cloud $1.19.
For a month of one RTX 4090, 730 hours at RunPod’s Community Cloud rate lists at $248.20 ($0.34 × 730), and at Vast.ai’s floor $321.78 ($0.4408 × 730), if the floor held for the month, which a floor need not do. Nothing here is forecast: these are the figures of one read.
How to check these figures yourself
Every figure above is listed under this text with the page it links and the time we read it. RunPod’s rates are on its pricing page; Vast.ai’s floors are taken from its public offer board. The RunPod vs Vast.ai page shows all 23 rows, each with its link and capture time, and updates with every hourly read; this article keeps the figures of October 19. The medians are on the RTX 4090 and A100 80GB card pages.