[Cursor Auto Model Selection: Satisfaction Close to Fable, Costs 60% Lower] According to monitoring by Beating Insights, Cursor has launched Cursor Router, designed for Teams and Enterprise to automatically select programming models. It first determines the task type and difficulty before deciding whether to invoke a cost-effective model or a cutting-edge model. The Router was trained on over 600,000 real requests and tested across millions of requests. Auto Intelligence's user satisfaction is close to Fable, with costs approximately 60% lower. Auto Balance's satisfaction surpasses Opus 4.8, with costs about 36% lower. Cursor has not disclosed the full model pool and does not default to showing which model was invoked for each request. The official statement confirms that the Router assigns models based on task type, such as delegating interface design to models better at visual tasks and assigning complex, lengthy tasks to models with stronger reasoning capabilities. [Original Link]