The Year in AI Papers
The year/Independent research

Paper 2604.27393

MiniCPM-o 4.5: Towards Real-Time Full-Duplex Omni-Modal Interaction

Junbo Cui, Bokai Xu, Chongyi Wang, Tianyu Yu, Weiyue Sun, Yingjing Xu, Tianran Wang, Zhihui He, Wenshuo Ma, Tianchi Cai, Jiancheng Gui, Luoyuan Zhang, et al.

Published
Apr 2026
Research lab
Independent
Citations
16
GitHub
26K stars

01 In brief

Summary

MiniCPM-o 4.5 is a 9B-parameter open-source multimodal large language model (MLLM) designed for real-time full-duplex omni-modal interaction, enabling simultaneous perception and response.

It introduces Omni-Flow, a unified streaming framework that aligns multimodal inputs and outputs along a shared temporal axis, converting turn-based interaction into a continuous, time-aligned process.

The model supports proactive behaviors like issuing reminders based on live scene understanding.

It approaches Gemini 2.5 Flash in vision-language capabilities, achieving state-of-the-art open-source performance at its scale, and surpasses Qwen3-Omni-30B-A3B in omni-modal understanding and speech generation quality.

The architecture includes multimodal encoders, an LLM backbone (Qwen3-8B), and speech decoders, all connected end-to-end.

Training involves staged pretraining, joint supervised fine-tuning, and reinforcement learning.

The model runs efficiently on edge devices with less than 12GB RAM, using an optimized llama.cpp-omni framework.

Evaluations show strong performance across vision-language, speech, text, and omni-modal benchmarks, with a win rate of 54.4 on LiveSports-3K-CC for full-duplex interaction.

Limitations include occasional speech instability and simple proactive behavior.

02 From the paper

Abstract

Recent progress in multimodal large language models (MLLMs) has brought AI capabilities from static offline data processing to real-time streaming interaction, yet they still remain far from human-level multimodal interaction. The key bottlenecks are no longer modality coverage or latency alone, but the interaction paradigm itself. First, perception and response are still separated into alternating phases, preventing models from incorporating new inputs for timely adjustment during generation. Second, most current models remain reactive, responding only to explicit user requests instead of acting proactively in the evolving multimodal environment. We present MiniCPM-o 4.5, our latest effort towards human-like multimodal interaction, which mitigates these gaps by real-time full-duplex omni-modal interaction. It can see, listen, and speak simultaneously in real-time, while also exhibiting proactive behaviors such as issuing reminders or comments based on its continuous understanding of the live scene. The key technique behind MiniCPM-o 4.5 is Omni-Flow, a unified streaming framework that aligns omni-modal inputs and outputs along a shared temporal axis. This formulation converts conventional turn-based interaction into a full-duplex, time-aligned process, enabling simultaneous perception and response and allowing proactive behavior to arise within the same framework. With a total of 9B parameters, MiniCPM-o 4.5 approaches Gemini 2.5 Flash in vision-language capabilities, delivering state-of-the-art open-source performance at its scale. It also surpasses Qwen3-Omni-30B-A3B in omni-modal understanding and delivers better speech generation, with significantly higher computation efficiency. Driven by its efficient architecture design and inference optimization, the model can perform real-time full-duplex omni-modal interaction on edge devices with less than 12GB RAM cost.