On October 14th, OpenRouter, an AI model aggregation platform, announced the launch of its private endpoint feature. This feature allows enterprises to connect their finely tuned AI models, dedicated model deployments, and reserved computing power to OpenRouter, using the same API as publicly available models on the platform. Enterprises can prioritize using their own deployed computing power, and when dedicated resources become unavailable or reach rate limits, they can switch to public model service providers according to preset rules, without needing to develop additional interfaces. Private endpoints are only visible to authorized organizations and will not appear in public model directories, leaderboards, or benchmark tests. This feature also supports cost statistics, automated API management, and data processing for specific regions. However, the connected models must be listed in the OpenRouter directory, and the deployment service provider must support providing its own API key (BYOK). Currently, this feature is only available to enterprise customers, with a maximum of 10 private endpoints configured per organization by default.
