Meta Develops Internal AI Router 'Switchboard' Targeting OpenRouter

According to reports citing internal documents, Meta's AAI Labs incubator organization has developed a competitive service to OpenRouter. Internal memos from July state AAI Labs has approved around 200 projects spanning consumer products, developer tools, and internal infrastructure. One such project is the AI model router 'Switchboard,' which scores user or AI agent request complexity to determine which model to deploy. Simple requests are directed to smaller, cheaper models, similar to OpenRouter's 'Auto Router.' The Switchboard project is still in early stages and may not become a product. Meta's team indicated Switchboard could be used for internal cost savings or shared with large-scale AI coding agent operations. Meta is expected to spend $14.5 billion on AI infrastructure this year, a more than double increase from last year. The model routing market has gained attention with entries from OpenAI, Databricks, and Palantir. OpenRouter, valued at $1.3 billion as of April, has recently seen discussions about potential acquisition by big tech firms.