I built a plan using Opus 4.5, and I want to build individual To-dos with Sonnet 4.5. However, when I build the To-do, Opus 4.5 is used. I’m unsure if this is because Opus was the model used to create the plan, and I won’t test other combinations, due to tokens being wasted.
Steps to Reproduce
Build a plan using one model (i.e. Opus 4.5)
At the top of the plan document, change the “Model used to build this plan” selection to something else (i.e. Sonnet 4.5).
At the bottom of the plan, click the bubble next to one or more To-Do list item(s) to select those item(s).
Click the “Build in New Agent” button at the right of the To-Do section header.
the selected To-Do(s) are built in a different model (in my case, Opus 4.5)
Expected Behavior
The To-Do(s) Build in New Agent using the selected Model used to build this plan.
This looks like a bug. The model selector in the plan should apply to “Build in New Agent”. I’ve passed it to the team.
For now, the workaround is to create the plan with the model you want to use for the build from the start, or build the whole plan instead. In that case, the model should be taken from the selector.
Let me know if you find any other patterns, like when it works and when it doesn’t.
Actually it looks like when you select a To-do and Build in New Agent, it is really using the model used by the most recently run agent, just like when you click the New Agent button in the Agents panel. Except unlike New Agent where you have the option to switch the Agent before entering the prompt, Build in New Agent runs automatically.
So if you select a To-do and Build in New Agent immediately after the plan was created, the model used to create the plan will also be used to Build the To-do, regardless of the selected “Model used to build this plan” at the top of the plan. This is what I had previously observed. But if you run another chat with a different model than what was used to write the plan, and then select a To-do and Build in New Agent, the model of that most recently run model will be used to build the to-do, regardless of the selected “Model used to build this plan,” as well as the model used to write the plan.
So I just noticed that if you click the New Agent button and switch the model in that new chat without needing enter a prompt, then you can select a To-do and click the Build in New Agent button, and it will use the model that you switched to in the new chat.
This is another indication of my theory that building a single todo isn’t using the same model selector that is used by the button which builds the whole plan. And once you know this, the user can use the workaround to select whichever model they want. It just isn’t as convenient as having a dropdown in the plan document itself.
Thanks for the info. This bug should already be fixed. Please update Cursor to the latest version (Help > Check for Updates) and check again. Now “Build in New Agent” should use the model selected in the plan selector.
Let me know if the issue is still there after updating.