RouterPlus is an AI inference platform for teams building applications with multiple AI models. It brings closed and open-weight models from different providers together behind one API key, helping teams compare options and route requests to models suited to their specific workloads. Teams can use RouterPlus credits or connect their own provider accounts, and compatible OpenAI- and Anthropic-based clients can get started by changing their base URL and key.

The platform evaluates models against a team’s own test cases and production traffic rather than relying solely on general benchmarks. Its Optimize feature compares candidates with a baseline on held-out evaluation cases and identifies a better option only when it meets quality and cost checks; teams decide whether to deploy changes. RouterPlus also offers regional routing, data-compliance controls, production-grade service-level agreements, and reserved capacity for anticipated inference needs. The gateway does not store prompts or completions, though it records usage metadata; providers apply their own data policies.