StepFun released the StepAudio 3 series of voice models, including five models now available on its open platform, with several ranking first globally on Artificial Analysis benchmarks.