Allow model selection to be stored per queued prompt (not applied globally)

Ideas Feature Requests
1

Feature request for product/service

Chat

Describe the request

I’d like to suggest a feature related to how queued prompts work in the Cursor chat workflow.

Current Behavior

Right now in Cursor, it’s possible to enqueue multiple prompts sequentially in the chat interface. For example:

  • I enter Prompt A
  • While Prompt A is running, I enqueue Prompt B
  • Then Prompt C
  • They execute sequentially in the order queued

However, the model selection behavior is currently global.

Whichever model is selected at the time execution begins is applied to all queued prompts, even if different models were selected when each prompt was originally enqueued.

In other words:

If I:

  1. Select Model A
  2. Enqueue Prompt A
  3. Switch to Model B
  4. Enqueue Prompt B

Both prompts end up running with Model B.

This makes it impossible to intentionally sequence different models for different tasks in a single queued workflow.

Why This Matters

Many real workflows benefit from running different prompts with different models.

For example:

  • Use a fast / cheap model for planning or scaffolding
  • Use a strong reasoning model for architecture or refactoring
  • Use a code-focused model for implementation
  • Use a small model for formatting or cleanup

This pattern is common in modern AI-assisted development pipelines.

Right now, to achieve this behavior, users must:

  • Wait for each prompt to finish
  • Manually switch models
  • Then send the next prompt

This breaks batching, slows workflows, and reduces the usefulness of the queue system.

Proposed Solution

Each queued prompt should store the model that was selected at the time the prompt was enqueued.

So the behavior would be:

  • Prompt A runs with Model A
  • Prompt B runs with Model B
  • Prompt C runs with Model C

independently.

This would make the queue behave more like a task scheduler, where each job carries its own execution configuration.

Impact

This change would enable:

  • True multi-model workflows
  • Faster batching of development tasks
  • More efficient token usage
  • Better control over cost vs performance
  • More advanced automation patterns

It would effectively turn the queue into a lightweight orchestration system for AI development.