The Dashboard → Playground lets you try catalog models in the browser. It uses your API key and calls the live Gateway Proxy, so every successful request is routed, logged, and billed the same way as an SDK call.
Playground is not the Admin Simulator. It does not use the master key and does not skip billing.
What you need
- A verified OctaRouter account and prepaid balance.
- At least one API key (Dashboard → API keys). The full secret is loaded only in memory for the playground session.
- Pick a model, or open Try in Playground from the Models page (
?model=).
Model kinds
The playground switches UI from the catalog (modalities + pricing_profile), not from a model_type column:
| Kind | Endpoint | Typical billing |
|---|---|---|
| Chat | POST https://api.octarouter.com/v1/chat/completions | LLM tokens ($/1M, including cache / reasoning when present) |
| Image | POST https://api.octarouter.com/v1/images/generations | Image tokens or per image (n=1) |
| Transcribe | POST https://api.octarouter.com/v1/audio/transcriptions | Per second or audio tokens |
| Speech | POST https://api.octarouter.com/v1/audio/speech | Per upstream character — never estimated from the text box |
Vision chat models still use the Chat panel and token billing. They are not image-generation models.
After you run a request
The catalog rate on the model is a list price. The amount shown after the run is charged_cost from request logs (route factor / schedule may apply). Cancelled or failed upstream calls are typically zero charge.
If the gateway returns a budget error, top up and retry.
See API protocols for the same endpoints from curl or an SDK.