Wallfacer Intelligence has open-sourced its next-generation full-modal flagship model, MiniCPM-o 4.5.

Wallfacer Intelligence has released its next-generation open-source multimodal model, MiniCPM-o 4.5. This model possesses full-duplex interaction capabilities, enabling users to "see, hear, and speak simultaneously," upgrading human-computer interaction from turn-based to real-time, free-flowing dialogue. Achieving state-of-the-art (SOTA) performance across all modalities through 9B parameters, it excels in visual understanding, document parsing, and speech generation, supporting real-time perception and proactive response.