Choose which model answers in Chat

In NeuralVault 2.0 the model button on the chat screen opens a picker with search, filters and a Runs on column that says where each model runs.

Last checked against the app on · View as Markdown

Steps

  1. Open a chat and press the model button. It shows the provider and model that will answer.
  2. Use the search box, or the filters "Sees images", "Recommended" and "Favourites". A simple view and an advanced view are available.
  3. Read the columns. They include Model, Context, Price per 1M and Runs on, and the advanced view adds more, such as Released and AI score.
  4. Check "Runs on". It says where the model runs: On this PC, OpenRouter, Anthropic, Local gateway, NeuralVault Managed, NeuralVault routing or My own keys.
  5. Choose a row. The model button now shows your choice, and later turns use it.

How to tell it worked

The model button at the top of the chat names the model you chose, and the next reply comes from it.

If it does not work

The list is empty, or the button says the model list could not be read.
The button’s hover text gives the reason. If the engine banner is showing, see the guide on the engine not running.
A model is listed but cannot answer.
The picker marks a model that cannot answer a chat turn and gives its reason on the row. Choose another.
The provider you want is not offered as usable.
Add its key first (see the guide on pasting an API key) or start Ollama for a local model.