VTX Macro Adds Custom Model Cascades for Every Profile

VTX Macro’s AI page introduces model cascades: ordered lists that let you choose your primary model and the alternatives VTX can try when a request fails. Each profile has independent lists for Main, Review, and Screener, so you can tailor the choices to each task.

Every model row includes the full model selector and its own supported parameters and timeout. You can add alternatives, move them into your preferred order, or remove them, with settings staying attached to each model as you rearrange the list. After an eligible failure, such as a rate limit, VTX can advance to the next saved model. Uncertain request outcomes and partially received responses stop the cascade; connected external models support automatic model fallback in Provider mode.

The update also makes Screener easier to use alongside Trade. Navigation fixes prevent overlapping runs when moving between pages, while a separate “Use my averages” control lets you view Screener cost estimates from the profile’s usage history independently of Main and Review. External model details reserve space while loading and wrap long host names more cleanly on smaller screens.

Runtime feedback is clearer, too. Client Mode recovery respects retry delays and request cooldowns. In Server Mode, when changed AI settings supersede an external analysis, an informational message explains that the next cycle uses the updated settings.

Browser-native WebGPU inference is removed in this release. If your profile uses WebGPU, select another model in AI settings. Local AI remains supported for running models through a configured local server.