Skip to main content
Every AI coworker runs on an AI model. By default Vybe picks the model for you with Auto. You can also pin a coworker to a tier or to a specific model, and set how hard that model thinks with its effort level.

​
Where you choose the model

A coworker has one model setting, shared by everyone who works with it. If you change the model in a coworker’s chat, you change it for everyone, in every conversation with that coworker.
Switching models doesn’t clear any history. The next reply uses the new model.

​
Auto

Auto (“Automatically pick the best model for the task.”) is on for new coworkers. It’s the toggle in the model menu. When a conversation starts, Auto looks at the request, decides what kind of work it is, and runs it on one of three tiers: While Auto decides, the model pill reads Choosing…. It then shows one, two or three bars for the tier it picked. Auto keeps the model it picked for the rest of that conversation. If a thread turns into harder work than it started as, start a new thread or pick a model yourself. For scheduled tasks and triggers set to Auto, it decides from the task description. Which model runs each tier, and which kinds of work go where, can change as Vybe updates its defaults. Owners and Admins can customize both for their organization, and add their own rules in plain language. See Models and gateway.

​
Pinning a tier

Turn Auto off in the model menu and the coworker moves to Balanced. The menu then lists Cheap, Balanced and Smart, each with the model it currently uses. Pick a tier to keep a coworker at that level for every conversation. The coworker uses whichever model your organization’s tier points to, so it follows any change to that tier.

​
Picking a specific model

Open Other models to see every model you can pick. Once a model is selected, click its name at the top of the menu to see the list again.
  • Use Sort by to order the list by Provider, Cost, Intelligence or Speed.
  • Hover over a model to see what it’s good for, who makes it, its effort levels, and bars for Cost, Intelligence and Speed. The cost line shows an estimate for a typical message, in credits and dollars.
  • Models marked Beta are newer: “Beta models are newer and may be less stable or change without notice. Use with caution.”
If a model is retired, coworkers set to it move to a replacement automatically.

​
Effort levels

Many models can think for longer before they answer. With a specific model selected, open Effort in the model menu and choose Low, Medium, High, Extra high or Max. Each model offers only the levels it supports, and its default is marked Default. Models with no effort setting show None for this model. Higher effort thinks longer and uses more output tokens, so replies take longer and cost more credits.

​
Choosing a model

  • Start with Auto. It sends simple work to cheaper models and hard work to smarter ones, which keeps costs down without you having to think about it.
  • Pin Smart, or a top model, for judgment-heavy coworkers, such as one that reviews contracts or plans strategy.
  • Pick a fast, low-cost model for simple, frequent jobs, such as a daily digest. Haiku 5.5, GPT-6 Luna, GPT 5.6 Luna and Gemini 3.5 Flash Lite are built for high-volume work.
  • Raise the effort before switching models when a good model is close but misses details.
  • Try a different model when a coworker is stuck. Different models have different strengths.
To see what your organization spends by model, use Settings → AI → Usage and split by model. See Plans and credits.

​
Available models

These are the models in the picker today. The list changes as new models come out and old ones retire, and the picker in Vybe is always current.

​
Anthropic

​
OpenAI

​
Google

​
Other

DeepSeek V4 Pro 0813 and DeepSeek V4 Flash 0731 can’t read images. Pick another model if your coworker needs to look at screenshots or photos.