Request to Restore Configurable Default Model for Dust Agents to Optimize Costs and Performance

To follow up on Marc-Antoine’s message, could you integrate the default model used by a Dust agent? This was possible before—why was this option removed? The default is now set to the standard model tier, which uses gptluna-high, even though gptluna-low is more than sufficient in terms of relevance for most of our everyday requests. Having the ability to configure the default model per agent would help us optimize performance and cost. We are a company of around 100 people, and since you changed your billing system to a token-based model, our invoice has increased almost sixfold. This is not sustainable for us in the long term. I would appreciate if you can make some adjuments like the feature we are asking.