What it is
A model picker on every agent, grouped by provider (Anthropic and OpenAI).
What it does
Lets you pick the model an agent runs on and see an estimated cost per run before you commit, so you can balance quality against spend.
Why it matters
You stay in control of the quality-vs-cost trade-off on every run, instead of paying flagship prices for simple jobs.
How it works
The agent runtime routes your chosen model through the AI gateway. The estimate multiplies the run's token budget by that model's per-token price.
How to use it
- Open an agent in the builder.
- Pick a model from the Model dropdown in the action bar.
- Check the estimated cost/run shown next to it, then Save or Run.