Chat playground
Test prompts and parameters, then copy the exact request into your code.
Start a conversation
Messages are sent to the selected model with the settings on the right.
Request code
This request reproduces the current conversation and settings.
cURL
About this playground
The chat playground sends your messages to the selected model through the same /v1/chat/completions endpoint your code will use, with the system prompt and sampling parameters from the panel. Every model here is open-weight and billed per token from your prepaid balance; the reply shows elapsed time and token usage so you can estimate cost before you integrate.
Tips
- Reasoning models (GPT-OSS, DeepSeek, GLM) think before answering — keep
max_tokensat a few hundred or more. - Use View code to copy a cURL request with exactly the parameters you set; swap in the Python or Node SDK from the model page.
- Compare speed and price on the comparison pages before choosing a default.