Research previewapply for access. We review each application and email you when it's approved, with your API key ready on the dashboard.

Workbench

The Workbench lets you prompt base models in your browser without writing code. Pick a model and type a raw prompt to generate a completion, with the same sampling controls as the API.

You can also explore completions in two views:

  • Loom — branch a prompt into a tree of completions: generate several continuations at once, pick one and keep going, backtrack to any point, and inspect per-token logprobs.
  • Compare mode — run one prompt across several models (or the same model at different settings) side by side.

Open Loom from the top nav. For Compare mode, use the Single / Compare toggle in any Workbench chat composer. Both are private to your account.