Step-Audio-R1.1, the native speech inference model of the Leap Star constellation, tops the leaderboard.

Step-Audio-R1.1, an open-source speech reasoning model from StepStar, topped the global rankings in the authoritative Artificial Analysis Speech Reasoning benchmark with an accuracy of 96.4%, surpassing mainstream models such as Grok and Gemini. It is the world's first open-source native speech reasoning model, possessing deep speech reasoning and real-time response capabilities, enabling end-to-end understanding of audio content and complex logical reasoning.