Playground
The Playground lets you test models interactively before integrating them into your application. It provides a chat-like interface with configurable parameters, so you can experiment with different models and settings without writing code.
Getting Started
- Navigate to Playground in the sidebar
- If this is your first time, the Playground will check that you have a model and API key selected
- If you don't have an API key yet, the Playground can automatically create one for you
Configuration Panel
On the left side (desktop) or in the slide-out drawer (mobile), you'll find the configuration panel.
Model Selection
Click the model dropdown to select which model to use. You can search by name to quickly find a specific model. The list includes all models available to your account.
API Key Selection
Select which API key to use for the request. Each key shows its name and assigned group. The Playground will create a temporary key named Playground-xxxxxx with unlimited quota if you don't have one.
System Prompt
Enter a system prompt to set the behavior and personality of the AI. This is equivalent to the system message in the Chat Completions format.
For example:
You are a helpful coding assistant. Always provide concise, working code examples.
Advanced Settings
Expand the Advanced Settings section to fine-tune model behavior:
| Parameter | Range | Default | Description |
|---|---|---|---|
| Temperature | 0 – 2 | 0.7 | Controls randomness. Lower = more deterministic, higher = more creative |
| Top P | 0 – 1 | 1 | Nucleus sampling — limits token selection to the top probability mass |
| Frequency Penalty | -2 – 2 | 0 | Reduces repetition of tokens that have already appeared |
| Presence Penalty | -2 – 2 | 0 | Encourages the model to talk about new topics |
| Max Tokens | 1 – 4096 | 2048 | Maximum number of tokens in the model's response |
Chatting with Models
Sending Messages
- Type your message in the input box at the bottom
- Press Enter to send, or Shift+Enter to add a new line
- The model responds with streaming output — you'll see the text appear in real time
Understanding Responses
- User messages appear on the right side with a blue background
- AI responses appear on the left side with a gray background
- AI responses support Markdown rendering, including code blocks, lists, tables, and more
Thinking Models
Some models (like DeepSeek R1, OpenAI o1, and o3) support an extended thinking mode. When these models reason through a problem, you'll see an amber Thinking section that you can expand or collapse. This shows the model's chain-of-thought reasoning before the final answer.
Clearing the Conversation
Click Clear Conversation in the configuration panel to reset the chat history and start fresh. This is useful when switching between different tasks or system prompts.
Tips for Effective Use
Finding the Right Temperature
| Temperature | Best For |
|---|---|
| 0 | Factual answers, code generation, data extraction |
| 0.3 – 0.7 | Balanced conversation, general Q&A |
| 0.7 – 1.0 | Creative writing, brainstorming |
| 1.0 – 2.0 | Experimental, highly varied outputs |
Using System Prompts Effectively
A good system prompt sets clear expectations:
- Define the role: "You are a senior Python developer..."
- Specify the format: "Respond in bullet points..."
- Set constraints: "Keep answers under 200 words..."
- Provide context: "The user is a beginner who..."
Testing Before Integrating
The Playground is the ideal place to:
- Compare how different models respond to the same prompt
- Tune temperature and other parameters for your use case
- Verify that a model supports the features you need (streaming, tools, vision)
- Test edge cases and error handling
Once you've found the right model and settings, check Quick Start to integrate them into your code.