Benchmarking models

Run a side-by-side

In the Playground, enter a representative prompt and run it across models at once. Compare answer quality, latency, and token cost, then set the winner as your agent's default — you can override per task later.