The models you can see and the models your agents actually use now come from the same published list, with real prices and real limits attached to each one.
That list also got bigger. You can connect four new providers alongside the ones already supported, and the Claude and Gemini defaults moved up to their current generation.
Changing an agent's model no longer means starting over. The conversation moves with it.
What you can do now
- See what a model costs and what its limits are before you pick it.
- Connect four new providers, and test a key the moment you add it.
- Change an agent's model in the middle of a conversation and keep the history.
- Switch straight from the message that tells you a usage cap was reached.
- Keep working when a model is unavailable, because another one picks it up.
Why it matters
Model choice used to be guesswork kept in two places at once. The list in front of you and the list agents could actually reach drifted apart, and neither told you what anything cost.
One list fixes the trust problem. Carrying the conversation across a switch fixes the practical one: you can try a cheaper or faster model on real work without losing what you already did.
Example workflows
- Budgets: A team moves routine drafting to a cheaper model and keeps the expensive one for review.
- Peak load: A capped model hands off mid-task and the work finishes anyway.
- Evaluation: A founder runs the same brief on two providers and compares the results.
- Operations: An admin adds a new provider key and confirms it works before the team relies on it.
What’s next
We're continuing to make the cost of a choice visible at the moment you make it.