How VibeMap picks a model

How VibeMap picks a model

You click Generate; VibeMap chooses the AI model for you. · Optional

You don't need to read this to use VibeMap. Come here if you're curious why a run used the model it did, or why you don't see a model picker on most buttons. Choosing the model by job size and plan (described below) is rolling out in stages. Until it reaches your workspace, VibeMap still picks the model for you, just without the size-based step.

What you do

Nothing. Press Generate (or type an instruction to the agent) and VibeMap selects the model for that job. There is no model dropdown to manage on the Generate button, the agent input, Prepare, schema generation, or the App Builder.

How the choice is made

Once size-based routing reaches your workspace, VibeMap looks at the job, not at a setting you have to remember:

  • What the job is. A schema, a system design spec, a batch of user stories, and a one-line edit each need different amounts of reasoning, so each starts from a different model.
  • How narrow it is. A small, focused edit (one field, one section, one story) can run on a lighter model. A button that works on the whole project keeps that job's usual model; it isn't upgraded just because the project is big.
  • How big your change is. When you type an instruction to the agent, a fast, lightweight classifier reads it first and estimates how large the change is. A small tweak stays on a lighter model; a sweeping rewrite is routed to a stronger one. This is the only case where VibeMap steps up to a stronger model than the job usually gets.

Your plan sets the ceiling

Once size-based routing reaches your workspace, your plan decides the strongest tier VibeMap is allowed to use for you. Routing never goes above it, and it can pick something lighter when the job doesn't need more.

PlanStrongest tier
Freeeconomy
Starterstandard
Pro and Teampremium
Enterprisefrontier (today this routes up to premium, since no frontier model is available yet)

Premium-tier runs draw on your premium token allowance. See Account and billing for how the monthly budget works.

Admins can still choose

Workspace admins can pick a specific model when testing. Everyone else gets VibeMap's choice, which keeps results consistent across a team.