One fleet, every workload
Training runs, fine-tunes, inference fleets, render farms and research campaigns — each needs a different shape of compute. Here is how to pick it.
Bring your framework — it already runs here
Most common
Train foundation models without waiting for a cluster
From a single 8-GPU node to multi-node runs on a 3.2 Tbps InfiniBand fabric — reserved when you need guaranteed capacity, on demand when you are still finding the recipe.
3.2Tbps
InfiniBand per node in a training cluster
8×H100
NVLink GPUs per HGX node
180GB
HBM3e on a single B200
Adapt open-weight models to your data in an afternoon
LoRA, full fine-tunes and preference tuning on a single GPU or one 8× node.
Serve models at the latency your users expect
Serve open-weight models on dedicated GPUs or autoscaling serverless endpoints.
Generate images and video without overpaying for VRAM
Diffusion and video models on GPUs sized for resolution, frames and latency.
Research compute that scales with the question
Life sciences, climate and simulation workloads on GPUs and dedicated metal.
A render farm that exists only while you render
GPU render farms on demand for animation, VFX and 3D pipelines.
Start with the right silicon
Our recommended starting points per workload. Live prices — billed per second, change instance type any time.
| Workload | Start on | Scale to | Start price |
|---|---|---|---|
| LLM training | H100 SXM ×8 Node | B200 SXM, H200 SXM | $19.92/hr |
| Fine-tuning & post-training | H100 SXM | H200 SXM, A100 80GB SXM | $2.69/hr |
| Production inference | L40S | H100 PCIe, RTX 5090 | $1.19/hr |
| Image & video generation | RTX 5090 | L40S, RTX 4090, H100 SXM | $1.29/hr |
| Scientific computing | A100 80GB SXM | H100 SXM, Metal E64 | $1.69/hr |
| Rendering & VFX | RTX 4090 | RTX 5090, L40S | $0.89/hr |
Start training in minutes
Create an account, add a card, and launch your first GPU instance. Per-second billing means you only pay for what you use.
No minimum commitment · Cancel anytime · $10 free credit for new accounts