# Model Neutrality and Routing ## The layer sells choice between models The vendor claims that it sends each task to the right model. Choice runs from frontier models to local models. The router balances quality, cost and privacy. The customer does not depend on one model vendor. > [!tip] Why this helps > Model prices fall and model rank changes often. A neutral layer turns that change into lower cost for the customer. ## Risks - Model vendors are also partners and possible competitors. - A router adds little value if one model is best at all tasks. - Local models have lower quality. The privacy gain must be worth it. ## Why it matters Neutrality is a strong story only if routing data improves results. Ask: does the router learn from outcomes? Without learning, routing is a feature, not a moat. See [[AI era Defensibility]]. ## Related - [[AI Deployment Layer MOC]] - [[Incumbent Bundling Risk]]