The Simplest Way for Humans and Agents to Use compute.
Across providers, SKUs, and workloads.
Quote a GPU. Run a function. Get a receipt.
$
- estimaterate locked · confirmation required
- provisionfresh H100-SXM requested
- ready80 GB HBM3 available
- stdoutstep 1/3 · loss 0.8421
- result{"status":"ok","artifact":"result.json"}
- terminatedmachine gone · receipt final
Provider usage$0.80
Platform fee$0.06
Total debit$0.86
Providers
- RunpodH100-SXMLive
- Hot AisleMI300X · 192 GBLive
- Vast.aiSpotLive · spot
- AWSTrainiumSoon
- Google CloudTPUSoon
First run.
Sign up, add credits, install the CLI, then run a function.
- 01
Sign up
- 02
Buy credits
$10 minimum.
- 03
Install the CLI
curl -fsSL https://compute.cx/install.sh | sh
- 04
Run a function
compute run simple_mlp.py::train --gpu H100
Guides.
Fine-tune a model on your data
- What
- Train an open model on your data.
- When
- When the base model misses your format or domain.
- Run
Small public model. Metrics only — adapters are not downloaded.compute run mi300x_lora_finetune.py::finetune --gpu MI300X
Train from a reward signal
- What
- Improve a model you can score.
- When
- Ranking, reward models, reasoning policies.
- Run
Stay on --wait or read logs later.compute run rl.py::train --gpu MI300X
Score a dataset overnight
- What
- Run a model over a file.
- When
- Evals, embeddings, labels, synthetic data.
- Run
Billed while the machine exists.compute run batch_infer.py::generate --gpu MI300X
Pass-through. Plus a simple fee.
You pay the provider's locked rate. We add a simple fee. A run that never becomes ready is $0.
See pricing