One API for speech-to-text, language models, and text-to-speech, with public benchmarks beside runtime availability.