Observed arrival · 2026-10-13
NomadGPU: GPU jobs with a recovery plan
A managed layer for running training and inference on Vast.ai GPUs, with host selection, checkpoint syncing, recovery after host failure, and spending limits.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Researchers and ML developers running jobs on rented community GPUs
- Worth noticing
- The page discloses that its AWS speed adjustment is estimated, not measured, and that AWS’s cited p4d instance has eight GPUs.
Field notes
Host selection is described as ranking reliability, network, PCIe and disk speed alongside traffic-inclusive hourly cost; an earlier run reportedly reached $0.81 when traffic was expensive. The published QLoRA example includes setup details—a PyTorch image, roughly 15 GB model download, 4-bit QLoRA and checkpoints every 50 steps—and labels AWS speed as estimated rather than measured.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
⊠LoginAccess appeared gated
$PaidCommerce or pricing visible
✦PrettyNotable craft visible
◎NicheUnusually specific use
One card from the complete issue