Speko is a single API that routes requests across speech-to-text, language and text-to-speech models from multiple providers. It selects models per use case and language, and publishes benchmarks alongside runtime availability.