At least AA have a dedicated speech model index, of which this is now the leader: https://artificialanalysis.ai/speech-to-speech