Playground
Try any model from the dashboard, and what the Playground key is.
The Playground page on the dashboard lets you try any model your workspace can use: a chat for text models, and a form for image, video, speech, transcription and music models. Each run is a real request, billed to your credit at the model's published price, and appears in your usage, your request log and, for media, on the Jobs page.
The Playground key
The first time you open the Playground in a workspace, the dashboard creates an API key for you
named Playground · <your name>. Every run you make in that workspace uses it.
| One per | member, per workspace |
| Monthly spend cap | $5.00 |
| Models | every model the workspace can use |
| Expires | never |
The key never reaches your browser. The dashboard keeps it encrypted and sends it to the API on your behalf, so a run in the Playground is admitted, routed and limited exactly like a request from your own code. Your workspace's policy, retention class and rate limits apply to it unchanged.
You open the same key from any browser or device you sign in on. Other members of the workspace get a key of their own.
Changing or revoking it
The key is listed on the Keys page with your other keys. Raise its cap there if $5.00 a month is not enough for your testing.
Revoking it stops every Playground run in that workspace at once. The next time you open the Playground, the dashboard creates a new key with the same name and cap, and the audit log records both the revocation and the new key.
Show code
Show code on any run gives you the same request as curl, Python or TypeScript. The snippet
reads your key from $OP_API_KEY and never contains the Playground key: use a key you created
yourself on the Keys page.
History
Your conversation history stays in your browser, for the last 20 conversations in each workspace. It is not stored on our servers and does not follow you to another device. In a zero retention workspace the Playground keeps no history at all.