Model Playground

AVAILABLE

Test foundation models and virtual routing policies directly within the Developer Portal without writing client code.

  • Portal Location: Projects > [Your Project] > Operations > Model Playground (/projects/[slug]/model-playground)

What is the Model Playground? #

The Model Playground allows engineers and prompt designers to:

  • Test raw foundation models (gpt-4o, deepseek-chat, claude-3-5-sonnet) side-by-side with virtual routing aliases (smart, fast).
  • Fine-tune hyperparameters: temperature, top_p, frequency_penalty, presence_penalty, and max_tokens.
  • Simulate streaming performance and measure real-time Time-To-First-Token (TTFT) latency.
  • Export formatted cURL, TypeScript, or Python snippets based on your active playground settings.
Rendering diagram...

How to Use the Model Playground #

  1. In your project workspace, click Operations > Model Playground.
  2. Select a target model or routing policy from the top dropdown.
  3. Enter your System Instructions and User Prompt.
  4. Adjust temperature and generation limits in the right settings drawer.
  5. Click Run or press Ctrl + Enter.
  6. Click View Code to copy production-ready SDK snippets.