Language

The language models the SDK can call for text generation and any agent that reasons in text. Each one is documented with the numbers that decide a choice: how much it can read at once, what it costs per million tokens in and out, and whether it reasons, calls tools, or reads images. You rarely name one by hand. The generation actions carry a sensible default, and a model override swaps in a specific model when a task needs more context, lower cost, or a particular capability.