LLM playground
Stream any of 338 models with your own key. Nothing is proxied — requests go from this page straight to the API, and the cost of every reply is printed under it.
Model
Stored only in this browser's localStorage. Your key never leaves your browser except to call the API directly.
Parameters
System prompt
Sent as the first message of every request.
anthropic/claude-sonnet-4.5
Bring your own key
Paste an API key in the panel to start. It's stored in this browser only and sent to nothing but the endpoint you choose — this page has no backend.
Get a key in under a minute →An LLM playground for 338 models
Most playgrounds lock you into one lab. This one speaks the OpenAI wire format, so every model in thecatalog — Claude, GPT, Gemini, Llama, DeepSeek, Qwen and the rest of 52 labs — runs in the same chat with the same request shape. Switch models between messages and compare answers and cost side by side.
What you test here is what you ship: point your SDK atapi.openkey.ai/v1, keep the model id, and the exact same request runs in production at the provider price + 3%.
Honest about keys
This page has no backend. Your key sits in your browser's localStorage and requests go directly from your browser to the endpoint you pick — we never see, proxy, or log them. Don't paste keys into playgrounds that can't say that.
No OpenKey key yet? Switch the endpoint to any OpenAI-compatible base URL — an OpenRouter key works today since the API shape is identical. Orget an OpenKey key in under a minute.
Try an example
Click a prompt to load it into the playground above.