Compare GPU clouds / Vast.ai
Vast.ai
Open GPU marketplace with live host prices
Marketplace
Vast.ai is a marketplace where independent hosts and certified data center partners list GPUs at prices set by supply and demand, updated hourly and billed per second. On 2026-11-09 its lowest live offers included H100 SXM at $2.27 (median $2.87), H200 at $3.43 and B200 at $5.50 per GPU-hour, across about 20 GPU models. Reserved terms of 1, 3 or 6 months are advertised at up to 50% off.
Best for: Price discovery and the cheapest live single-node GPU offers.
ClusterMAX 3.0: Participation
- ✓ Publishes prices
- – Serverless inference
- – AMD GPUs
- – European data centers
1 Legal nameVast.ai
2 HeadquartersLos Angeles, CA, USA
3 Founded2016
4 Funding or listingNot disclosed
5 How you rentMarketplace, On-demand, Reserved clusters
6 Minimum commitmentNone (per-second billing, no minimum hours); optional reserved terms of 1, 3 or 6 months
7 InterconnectNot disclosed
8 OrchestrationVast CLI, REST API, Python SDK
9 GPUs offeredGB200, B300, B200, H200, H200 NVL, H100 SXM, H100 PCIe, H100 NVL, A100 80GB, L40S, L40, L4, RTX PRO 6000, RTX 6000 Ada, RTX 5090, RTX 4090, RTX 3090, A40, A10
10 RegionsGlobal, host-dependent (40+ data centers stated)
11 CertificationsSOC 2 Type I (2025)
12 Customers and partnersNot disclosed
Published prices
| GPU | Price | Basis | Source |
|---|---|---|---|
| RTX 4090 | $0.14 | marketplace, lowest live offer (median $0.49) | Pricing page |
| A100 80GB | $0.33 | marketplace, lowest live offer, PCIe (median $0.73) | Pricing page |
| A100 80GB | $0.37 | marketplace, lowest live offer, SXM4 (median $0.84) | Pricing page |
| RTX 5090 | $0.40 | marketplace, lowest live offer (median $0.67) | Pricing page |
| L40S | $0.47 | marketplace, lowest live offer (median $0.80) | Pricing page |
| RTX PRO 6000 | $1.06 | marketplace, lowest live offer, server edition (median $1.65) | Pricing page |
| H100 SXM | $2.27 | marketplace, lowest live offer (median $2.87) | Pricing page |
| H100 PCIe | $2.47 | marketplace, lowest live offer (median $2.47) | Pricing page |
| H100 NVL | $2.53 | marketplace, lowest live offer (median $2.68) | Pricing page |
| H200 NVL | $3.33 | marketplace, lowest live offer (median $4.00) | Pricing page |
| H200 | $3.43 | marketplace, lowest live offer (median $4.34) | Pricing page |
| B200 | $5.50 | marketplace, lowest live offer (median $8.75) | Pricing page |
| GB200 | $8.13 | marketplace, lowest live offer (median $9.06) | Pricing page |
Where we recommend them
- Cheapest published H100, an alternative.
- Best marketplace, an alternative.
Worth knowing before you sign
- Prices move hourly and host quality varies; the pricing card does not say whether the lowest offer is on-demand or interruptible.
- States SOC 2 Type I only.
- In the Participation tier of ClusterMAX 3.0; SemiAnalysis called the UI hard to use.
Sources
Work at Vast.ai? Send corrections with a source to editors@bestneoclouds.com. We fix verified errors within five business days.
Compare with
- Prime Intellect GPU compute marketplace with its own clusters and post-training tools
- Shadeform GPU aggregator across about 18 clouds at the clouds’ own prices
- Hyperbolic GPU marketplace of supplier capacity, plus multi-cloud GPU software (Forge)
- SF Compute Order-book market for GPU cluster time, with resale
- Mithril Bid-based GPU spot market, formerly Foundry