01
GPUs
Every GPU, one rate
The four GPUs people rent most, out of 38 on the live market. Each one is billed by the second from your USDG balance.
Meridian Health
80 GB VRAM
2026
Training and fine-tuning
The workhorse of AI labs. Fine-tune a model on your own data, train from scratch or serve a large model with room to spare.
$2.96
Per GPU hour, billed by the second
Meridian Health
80 GB VRAM
2026
Training and fine-tuning
The workhorse of AI labs. Fine-tune a model on your own data, train from scratch or serve a large model with room to spare.
$2.96
Per GPU hour, billed by the second
Meridian Health
80 GB VRAM
2026
Training and fine-tuning
The workhorse of AI labs. Fine-tune a model on your own data, train from scratch or serve a large model with room to spare.
$2.96
Per GPU hour, billed by the second
Carrow Logistics
180 GB VRAM
2026
The biggest models
NVIDIA’s latest flagship. Enough memory to serve a 70B model on a single card, and the fastest training you can rent.
$6.58
Per GPU hour, billed by the second
Carrow Logistics
180 GB VRAM
2026
The biggest models
NVIDIA’s latest flagship. Enough memory to serve a 70B model on a single card, and the fastest training you can rent.
$6.58
Per GPU hour, billed by the second
Carrow Logistics
180 GB VRAM
2026
The biggest models
NVIDIA’s latest flagship. Enough memory to serve a 70B model on a single card, and the fastest training you can rent.
$6.58
Per GPU hour, billed by the second
Veldt Insurance
24 GB VRAM
2025
Images, video and small models
The cheapest way to run Stable Diffusion, FLUX, ComfyUI or a 7B model. Fast, small and priced like a coffee.
$0.37
Per GPU hour, billed by the second
Veldt Insurance
24 GB VRAM
2025
Images, video and small models
The cheapest way to run Stable Diffusion, FLUX, ComfyUI or a 7B model. Fast, small and priced like a coffee.
$0.37
Per GPU hour, billed by the second
Veldt Insurance
24 GB VRAM
2025
Images, video and small models
The cheapest way to run Stable Diffusion, FLUX, ComfyUI or a 7B model. Fast, small and priced like a coffee.
$0.37
Per GPU hour, billed by the second
H200 SXM
141 GB VRAM
2025
Long contexts and big batches
An H100 with almost twice the memory. Serve bigger models, longer contexts and larger batches without splitting across cards.
$3.95
Per GPU hour, billed by the second
H200 SXM
141 GB VRAM
2025
Long contexts and big batches
An H100 with almost twice the memory. Serve bigger models, longer contexts and larger batches without splitting across cards.
$3.95
Per GPU hour, billed by the second
H200 SXM
141 GB VRAM
2025
Long contexts and big batches
An H100 with almost twice the memory. Serve bigger models, longer contexts and larger batches without splitting across cards.
$3.95
Per GPU hour, billed by the second