Playground
Try any model on the shelf. Every answer is a real request, billed to your balance at the price on its card.
GPT-5 nano$0.05 in / $0.40 out
Pick a model and send a message. This is the real gateway, not a demo.
Model parameters
How long the model thinks before it answers. Thinking costs more: those tokens are billed at the output rate.
This model has no temperature setting.
Context400,000 tokens
Price$0.05 in / $0.40 out per 1M tokens