Inference Playground

Test inference models interactively in your browser before integrating into your applications.

Accessing the Playground

1. Navigate to Inference > Playground
2. Select a model from the dropdown
3. Start sending messages

Features

Chat Interface

• Send messages and receive model responses
• View full conversation history
• Clear conversation to start fresh
• Real-time streaming responses

Model Selection

Choose from any available model in the catalog:

• Switch models mid-conversation
• Compare responses across models
• View model details and pricing

Parameter Tuning

Adjust inference parameters in real-time:

Token Usage

The playground tracks token usage per conversation:

• Input tokens per message
• Output tokens per response
• Total tokens consumed
• Estimated cost

Performance Metrics

View real-time performance:

• Time to first token
• Tokens per second
• Total response time

Use Cases

Prompt Engineering

Test and refine prompts before deployment:

System: You are a technical documentation writer.
User: Write a brief explanation of GPU memory bandwidth.

Model Evaluation

Compare models for your specific use case:

1. Send the same prompt to different models
2. Compare response quality
3. Compare speed and cost
4. Choose the best model for your needs

Quick Prototyping

Test model capabilities before writing integration code:

• Verify the model understands your domain
• Test edge cases
• Validate output format

Limitations

• Playground uses your inference API quota
• Conversations are not persisted between sessions
• Large responses may be truncated at the model’s context limit

Next Steps

• Generate API Keys to integrate into your application
• Browse the Model Catalog for more models
• View API Reference for programmatic access