Vast.ai Alternatives: GPU Cloud Hosts
Vast.ai alternatives verdict: Stay on Vast.ai when you want a GPU marketplace whose official public pricing feed (updated 11 November 2026 at 14:30:20 UTC, 15:30 Europe/Paris) listed a median of $0.494 per hour for an RTX 4090 (minimum $0.1343, 505 offers) and $2.3385 per hour for an H100 SXM (minimum $2.1338, 24 offers), cited here with attribution to Vast.ai. SaladCloud publishes a fixed public grid billed per second, with an RTX 4090 at $0.440 / $0.347 / $0.253 / $0.160 per hour across High, Medium, Low and Lowest priority (official SaladCloud pricing, same day). Runpod lists an RTX 4090 at $0.89 per hour and an H100 SXM at $3.99 per hour on its Secure Cloud card (official Runpod pricing, same day). A GPU-by-GPU Runpod comparison already lives on our Vast.ai vs Runpod page; this shortlist is the wider alternatives view. Lambda is listed as a GPU-cloud alternative only: its public pricing page does not disclose current on-demand rates without a logged-in account, so this article quotes no Lambda dollar amounts.
Stay on Vast.ai if: you want the cheapest available offer on a given GPU class (the feed’s minimum on an RTX 4090 was $0.1343 per hour), you can accept host-to-host variation, or you plan to convert an on-demand instance to reserved pricing for up to 50% off, as Vast.ai docs describe.
Look at an alternative if: you want a published, non-marketplace list price (Runpod or SaladCloud), per-second billing with priority tiers (SaladCloud), or a serverless, per-second GPU API (Modal).
- Vast.ai: marketplace medians and minima from the official public feed, 11 November 2026, 14:30:20 UTC; interruptible instances often 50%+ cheaper than on-demand; reserved conversions up to 50% off (Vast.ai docs).
- SaladCloud: from $0.015 per GPU-hour at Lowest priority on additional classes; RTX 4090 $0.440 to $0.160; RTX 5090 $0.600 to $0.250; billed per second; vCPU and RAM included on GPU instances.
- Runpod Secure Cloud: H100 SXM $3.99/hr, H100 PCIe $2.89/hr, A100 PCIe or SXM $1.79/hr, B200 $7.99/hr, H200 $5.29/hr, RTX 4090 $0.89/hr, RTX 5090 $1.19/hr.
By the Unfair Advantage editorial team · Updated November 11, 2026. Method: reading of the official Vast.ai pricing page, Vast.ai pricing.md (agent pricing notes), the official public GPU pricing feed at storage.googleapis.com/vast-public-gpu-pricing/gpu-pricing-public.json (feed stamp 2026-11-11T14:30:20Z), Vast.ai instance pricing docs, the official SaladCloud, Runpod, Modal and Lambda pricing pages, all fetched November 11, 2026 (Europe/Paris). We did not rent instances or open paid accounts. Competitor prices are quoted as text only, with source URLs in the plain-text source list. Live marketplace medians move; treat the feed stamp as the quote date. For storage, bandwidth and host-trade detail, see our live Vast.ai pricing page and the live Vast.ai review.
When should you stay on Vast.ai, and when should you switch?
Stay when a marketplace bid — including a low minimum offer — is the point, and you can pick hosts by price, reliability and region. Price a second cloud when you need a published list price, a priority tier with per-second billing, or a serverless API that hides host selection.
Stay on Vast.ai when…
- You want the official feed’s minimums, such as $0.1343 per hour on an RTX 4090 or $2.1338 on an H100 SXM (15:30 Europe/Paris, 11 November 2026).
- You can work with interruptible instances, which Vast.ai docs say are often 50%+ cheaper than on-demand.
- You plan to convert on-demand rentals to reserved pricing for up to 50% off, as the same docs describe.
- You already compared Runpod in detail on our live vs page and still prefer marketplace spreads.
Price an alternative when…
- You want a public, non-auction list: Runpod’s Secure Cloud card or SaladCloud’s High-to-Lowest grid.
- You want per-second billing and four priority tiers, which SaladCloud publishes alongside monthly Lowest columns.
- You want a per-second GPU API. Modal lists an Nvidia H100 SXM5 at $0.001097 per second.
- You want a reserved or single-operator cloud such as Lambda, knowing that current public rates are not disclosed without a logged-in account, so no Lambda dollar amounts appear here.
What does the official Vast.ai pricing feed list?
Vast.ai does not publish a single fixed hourly menu on its marketing pricing page. The vendor’s pricing.md file points agents to a public JSON feed, free for display and citation with attribution. We fetched that feed on 11 November 2026; it is stamped 2026-11-11T14:30:20Z (15:30 Europe/Paris). Figures below are the feed’s current minimum, 10th percentile and median hourly prices, plus the offer count, cited with attribution to Vast.ai.
| GPU | Minimum $/hr | p10 $/hr | Median $/hr | Offers |
|---|---|---|---|---|
| RTX 4090 (24 GB) | 0.1343 | 0.3614 | 0.494 | 505 |
| RTX 5090 (32 GB) | 0.3874 | 0.4804 | 0.6685 | 712 |
| RTX 3090 (24 GB) | 0.0943 | 0.1481 | 0.203 | 213 |
| H100 SXM (80 GB) | 2.1338 | 2.1338 | 2.3385 | 24 |
| H100 PCIe (80 GB) | 2.0014 | 2.0014 | 2.5344 | 5 |
| A100 SXM4 (80 GB) | 0.4008 | 0.4015 | 0.8416 | 70 |
| A100 PCIe (80 GB) | 0.4019 | 0.4023 | 0.8069 | 51 |
| H200 | 2.1066 | 2.1066 | 4.3431 | 45 |
| B200 | 6.6731 | 6.6731 | 10.0003 | 42 |
The H100 PCIe row had only five offers at the stamp, so that median can move quickly. Storage is billed continuously while retained, and bandwidth is a separate transfer cost, according to Vast.ai instance pricing docs. Those add-ons are covered on our live Vast.ai pricing page.
How does SaladCloud compare as a Vast.ai alternative?
SaladCloud publishes a fixed public grid, billed per second, and states that GPUs start from $0.015 per hour. The High / Medium / Low / Lowest columns for consumer flagship cards, fetched 11 November 2026, are:
- RTX 5090 32 GB: $0.600 / $0.483 / $0.367 / $0.250 per hour; Lowest monthly column $182.50.
- RTX 4090 24 GB: $0.440 / $0.347 / $0.253 / $0.160 per hour; Lowest monthly column $116.80.
- RTX 3090 24 GB: $0.200 / $0.163 / $0.127 / $0.090 per hour; Lowest monthly column $65.70.
vCPU and RAM are included with every GPU instance on that page. CPU-only groups are billed separately at $0.005 per vCPU-hour and $0.001 per GB of RAM-hour. Higher priority can preempt lower-priority work. Salad Dedicated reserved capacity is quoted separately and is not in the public grid.
Compared with Vast.ai’s same-day median of $0.494 on an RTX 4090, SaladCloud’s High tier ($0.440) sits close, and Lowest ($0.160) undercuts the Vast.ai median — at the cost of being the most interruptible tier on SaladCloud’s own labels. Vast.ai’s minimum of $0.1343 can still undercut Salad Lowest, depending on the host.
How does Runpod compare as a Vast.ai alternative?
Runpod’s official Secure Cloud card (fetched 11 November 2026) lists per-hour prices, with a per-second toggle on the page. Selected rows:
- B300 $8.99/hr · B200 $7.99/hr · H200 $5.29/hr · H100 SXM $3.99/hr · H100 PCIe $2.89/hr · H100 NVL $3.19/hr.
- A100 PCIe and A100 SXM $1.79/hr · RTX 5090 $1.19/hr · RTX 4090 $0.89/hr · RTX 3090 $0.50/hr · RTX A6000 $0.59/hr.
Those list prices sit above Vast.ai’s same-day medians on the overlapping cards (H100 SXM median $2.3385 vs Runpod $3.99; RTX 4090 median $0.494 vs Runpod $0.89). You pay for a published rate and a single operator, not a host marketplace. The GPU-by-GPU storage and reliability comparison is already on our live Vast.ai vs Runpod page; this section only restates today’s Runpod list so the wider shortlist stays dated.
How do Modal and Lambda compare?
Modal bills GPU tasks per second. Official resource costs fetched 11 November 2026 include Nvidia H100 SXM5 at $0.001097 per second, Nvidia H200 SXM at $0.001261 per second, Nvidia B200 at $0.001736 per second, and Nvidia A100 80 GB at $0.000694 per second. UA arithmetic, labeled: $0.001097 × 3,600 = $3.9492 per hour for the H100 SXM5; $0.000694 × 3,600 = $2.4984 per hour for the A100 80 GB. Modal also lists a Starter plan at $0 plus compute per month, with $30 per month of free credits and 3 workspace seats, on the same page. That is a serverless shape, not a rented VM.
Lambda remains in this shortlist as a single-operator GPU cloud, but this article quotes no Lambda dollar amounts. The official public pricing page does not disclose current on-demand rates without a logged-in account, and we did not log in.
What does an RTX 4090 or H100 cost across these clouds?
Snapshot of official figures fetched 11 November 2026. Vast.ai columns are the public feed’s current minimum and median at 14:30:20 UTC. SaladCloud shows High and Lowest. Runpod is the Secure Cloud list. Modal is UA arithmetic from the official per-second rate, labeled. An em dash means the public page we fetched did not publish that card in the grid we quoted.
| GPU | Vast.ai min | Vast.ai median | Salad High / Lowest | Runpod list | Modal (UA $/hr) |
|---|---|---|---|---|---|
| RTX 4090 | $0.1343 | $0.494 | $0.440 / $0.160 | $0.89 | — |
| RTX 5090 | $0.3874 | $0.6685 | $0.600 / $0.250 | $1.19 | — |
| RTX 3090 | $0.0943 | $0.203 | $0.200 / $0.090 | $0.50 | — |
| H100 SXM | $2.1338 | $2.3385 | — | $3.99 | $3.9492 (SXM5) |
| A100 80 GB | $0.4019 (PCIe) | $0.8069 (PCIe) | — | $1.79 | $2.4984 |
| H200 | $2.1066 | $4.3431 | — | $5.29 | $4.5396 |
| B200 | $6.6731 | $10.0003 | — | $7.99 | $6.2496 |
Modal H200 UA arithmetic: $0.001261 × 3,600 = $4.5396 per hour. Modal B200: $0.001736 × 3,600 = $6.2496 per hour. Vast.ai’s B200 median ($10.0003) sits above Runpod’s $7.99 list, while Vast.ai’s minimum ($6.6731) sits close to Modal’s UA hourly figure. Always re-check the live feed before you rent.
How do billing models compare?
Vast.ai charges by the rented offer: GPU time plus continuous storage and bandwidth, with on-demand, interruptible and reserved options described in the docs. There is no flat “starter plan” on the marketing pricing page we fetched. SaladCloud bills per second of running state and states that allocation, download and cold start are free; monthly Lowest columns use a 730-hour convention on the grid we read. Runpod shows per-hour and per-second toggles on the Secure Cloud card. Modal bills per second of GPU, CPU, memory and volume time, plus network egress. Lambda’s current on-demand rates are not disclosed on the public pricing page without a logged-in account, so no Lambda figures appear in this billing comparison.
What pitfalls should buyers watch for?
- Do not treat a Vast.ai median as a guaranteed checkout price. The official feed is a snapshot of available offers and updates hourly.
- Thin rows move fast: the H100 PCIe median sat on only five offers at the stamp we cite.
- SaladCloud’s Lowest tier is the most interruptible on that vendor’s own labels; High can preempt it.
- Lambda’s public pricing page does not disclose current on-demand rates without a logged-in account. This article therefore quotes no Lambda dollar amounts.
- Modal’s per-second rates need a × 3,600 conversion before you compare them with hourly cards; we label that arithmetic as ours.
- Storage and bandwidth on Vast.ai are not inside the GPU median. See the live pricing page before you estimate a long job.
Which GPU cloud should you pick?
Pick Vast.ai when a live minimum or a mid-pack median on the official feed beats a competitor’s list — as it did on 11 November 2026 for an H100 SXM median of $2.3385 against Runpod’s $3.99 list, and for an RTX 4090 minimum of $0.1343 against SaladCloud Lowest at $0.160. Pick SaladCloud when you want a published priority grid and per-second billing on consumer GPUs. Pick Runpod when a single-operator Secure Cloud list matters more than a marketplace spread. Pick Modal for a per-second API with a $0 Starter plus $30 of monthly credits. Leave Lambda out of a same-day price bake-off unless you can read current rates inside a logged-in account; this article quotes no Lambda dollar amounts.
FAQ
Are Vast.ai GPU prices fixed?
No. Vast.ai is a marketplace. The official public pricing feed, stamped 11 November 2026 at 14:30:20 UTC (15:30 Europe/Paris), listed a median of $0.494 per hour for an RTX 4090 and $2.3385 for an H100 SXM, alongside cheaper minima. Rates move with inventory.
How does SaladCloud compare on an RTX 4090?
SaladCloud’s public grid lists an RTX 4090 at $0.440, $0.347, $0.253 and $0.160 per hour for High, Medium, Low and Lowest priority, billed per second, with vCPU and RAM included on GPU instances (official SaladCloud pricing, 11 November 2026).
Is Runpod cheaper than Vast.ai on an H100 SXM?
Not on the figures we fetched. Runpod’s Secure Cloud list shows $3.99 per hour for an H100 SXM, while Vast.ai’s official feed median the same day was $2.3385 and the minimum was $2.1338. You still may prefer Runpod for a single-operator list price.
Can I use Lambda’s website prices today?
Not as a public current list. The official Lambda pricing page does not disclose current on-demand rates without a logged-in account, which we did not open. This article therefore quotes no Lambda dollar amounts.
Where do these Vast.ai figures come from?
From Vast.ai’s official public GPU pricing feed, which the vendor’s pricing.md file licenses for display and citation with attribution. We fetched it on 11 November 2026; the file is stamped 2026-11-11T14:30:20Z. It is not a fixed catalog.
Sources
Sources are listed as plain URLs, without links, so this page sends no clicks to vendors other than through the disclosed affiliate button. Dates are Europe/Paris verification dates unless a source stamps its own date.
- https://vast.ai/pricing — official pricing page, fetched 11 Nov 2026, Europe/Paris.
- https://vast.ai/pricing.md — agent pricing notes and feed licence, fetched 11 Nov 2026.
- https://storage.googleapis.com/vast-public-gpu-pricing/gpu-pricing-public.json — official public feed, stamp 2026-11-11T14:30:20Z (15:30 Europe/Paris). Cited with attribution to Vast.ai.
- https://docs.vast.ai/guides/instances/pricing — instance types, interruptible and reserved notes, fetched 11 Nov 2026.
- https://vast.ai/terms — terms URL referenced by the feed, fetched 11 Nov 2026.
- https://salad.com/pricing — official SaladCloud pricing, fetched 11 Nov 2026.
- https://www.runpod.io/pricing — official Runpod pricing, fetched 11 Nov 2026.
- https://modal.com/pricing — official Modal pricing, fetched 11 Nov 2026.
- https://lambda.com/pricing — official Lambda pricing page, fetched 11 Nov 2026. Current on-demand rates are not disclosed without a logged-in account; no Lambda dollar amounts are quoted in this article.