模思智能
基础模型与学习范式
模思智能源于复旦大学语音研究团队，以 SpeechGPT 等成果探索能感知、表达并实时回应的语音模型。其方向是让语音智能体处理打断、情绪和自然对话，而不只是朗读文字。
Originating from Fudan speech research, Mosi builds on SpeechGPT to pursue voice models that perceive, generate and respond in real time. Its focus moves beyond reading text toward agents that handle interruptions and richer conversational cues.
