What is cpmGO
cpmGO (MiniCPM Super Assistant) is the world's first pure on-device intelligent assistant launched by ModelBest, designed specifically for automotive smart cockpits. cpmGO is developed based on ModelBest's MiniCPM on-device model, featuring rich capabilities such as vision, voice, multimodal interaction, and graphical UI interaction, enabling full-chain perception, decision-making, and execution from outside to inside the cockpit. cpmGO provides a "native on-device experience" aligned with cloud-based large models, supports weak-network or offline environments, protects user privacy, and responds quickly. cpmGO offers full-scenario shelf-ready atomic products, supports personalized customization, and helps smart cars achieve smarter, safer, and more efficient interactive experiences.
Main features of cpmGO
- Multimodal interaction: Supports multiple interaction methods such as voice, gestures, and graphical UI, enabling "what you see is what you can say."
- Intelligent decision-making and execution: Understands user intent and automatically completes tasks such as navigation and air conditioning adjustment.
- Privacy protection: Data processing is completed locally without relying on the cloud, ensuring user privacy and security.
- Suitable for weak-network environments: Supports stable operation in weak-network or offline environments, achieving low power consumption and high performance.
Technical principles of cpmGO
- On-device model architecture: Based on ModelBest's MiniCPM on-device model, the model is optimized to support efficient local operation while maintaining strong performance. To meet the needs of automotive cockpits, the model's functions and size are precisely tailored to match the requirements of automakers.
- Deep adaptation and inference optimization: Deeply adapted to mainstream automotive chips, with joint tuning and inference acceleration to solve the issues of power consumption and computing power usage of automotive chips. Based on optimized algorithms and hardware acceleration, it achieves millisecond-level response speed.
- Multimodal perception and interaction: Combines multiple modalities such as vision, voice, and graphical UI to achieve a more natural and efficient interactive experience. Based on deep learning algorithms, it understands users' vague intentions and provides precise services.
- Device-cloud collaboration: Based on device-cloud collaboration, it enables more powerful feature expansion and data updates.
Application scenarios of cpmGO
- Smart car users: Those pursuing a high-tech driving experience, conveniently controlling the in-car system based on intelligent interaction methods such as voice and gestures.
- Privacy-focused users: Users with high requirements for personal data privacy, hoping that in-car data will not be transmitted externally to ensure information security.
- People who often drive in weak-network environments: Such as those who often drive in tunnels, mountainous areas, or remote regions, and need to use in-car functions normally in weak-network or offline environments.
- Family users: Those who need an intelligent assistant to help care for children or pets in the car, identifying and reminding about situations such as children not wearing seat belts or abnormal pet conditions.
- Users pursuing an efficient driving experience: Enhancing the convenience and safety of operations during driving.