Tutorials
Step-by-step walkthroughs with real commands and real output - a 125B model on one GPU, training, sweeps, JupyterLab, and coding agents.
Each tutorial is one task from start to finish, with the commands as they were typed and the output they printed.
| Tutorial | What you do | Cost |
|---|---|---|
| Try a 125B model on a 96 GB GPU | Serve a large mixture-of-experts model and chat with it from your laptop | about $2.75 an hour |
| Train with uv scripts | Run a training script on a rented GPU and get its outputs back | GPU time |
| Run a parameter sweep | Run many trainings at once and compare them | GPU time |
| JupyterLab on a rented GPU | Open a notebook on a remote GPU in your browser | GPU time |
| Let your coding agent use a GPU | Give Claude Code or Codex the hi skill so they can rent GPUs safely | GPU time |