Cursor has made Cursor Router generally available for Teams and Enterprise plans. The system is a request-level classifier that inspects each prompt before execution and dispatches it to the most suitable model, aiming to lower costs without sacrificing quality.
- The classifier is trained on 600k+ live requests and optimized for user satisfaction (AFC) as its reward signal.
- It analyzes query, context, task complexity, and domain to route simple work to price-efficient models, UI updates to high-taste models, and complex problems to frontier reasoning models.
- Evaluation relies on online A/B tests across millions of requests rather than offline evals, accounting for cache-miss costs incurred when switching models mid-conversation.
- Auto Intelligence mode offers about 60% lower cost than Opus 4.8 while matching Fable's user satisfaction; Auto Balance provides comparable satisfaction to GPT-5.6 Sol at a lower spend rate.
- Measured cost per commit is $4.63 for Auto Balance and $6.76 for Auto Intelligence, compared to $7.34 for Opus 4.8.
Cursor Router addresses the inefficiency of developers using a single frontier model for all tasks by dynamically matching task difficulty to appropriate pricing tiers.