SenseTime’s Cantonese AI Assistant Sensechat Receives a Comprehensive Upgrade, Delivering a More Native Multimodal AI Experience for Hong Kong Users

SenseTime officially launched the enhanced version of Sensechat, its Cantonese AI assistant app, today. Built on the SenseNova V6 large model system, the latest version of Sensechat achieves a significant technical breakthrough in real-time audio and video interaction. Beyond “listening” and “speaking,” is now able to also “see” and “think”, supporting a wide range of scenarios. Hong Kong users can download or update the app for free on iOS and Android platforms to experience a smarter, more efficient, and more localized multimodal AI assistant.

New Real-Time Audio and Video Interactions for a More Localized AI Experience

• Functional Upgrades: The updated version introduces real-time audio and video interactions, enabling the AI assistant to not only “listen” and “speak” but to also “see” and “think.” Combined with deep reasoning and online search capabilities, Sensechat delivers a more accurate and seamless interactive experience. Currently, the audio and video calls support Mandarin, while Cantonese support will be released in a subsequent version.

• Emotional Interactions: The audio and video call feature includes emotional variation capabilities, allowing the assistant to detect and express emotions such as happiness, anger, or sadness. This simulates realistic emotional interactions and provides a more personalized user experience.

•  Enhanced Localization: Sensechat’s understanding of Hong Kong’s local culture, customs, and trending social topics has been further enhanced. It now fully supports Traditional Chinese conversations and better aligns with the linguistic preferences of native Hong Kong users.

Powered by SenseNova V6 for Advanced Multimodal Reasoning

With robust support from the SenseNova V6 large model, Sensechat features advanced multimodal and deep reasoning capabilities, enabling precise analysis and logical inference for complex tasks. As one of China’s leading native multimodal general purpose large models, SenseNova V6 is built on a Mixture-of-Experts (MoE) architecture with over 600 billion parameters and has achieved major breakthroughs in several key technical areas:

• Multimodal Long Chain-of-Thought: Trained on over 200 billion high-quality multimodal long-chain reasoning data, supporting thinking chain synthesis of up to 64K in length, greatly enhancing the model’s depth of reasoning.

• Mathematical Capabilities: Demonstrates significantly stronger performance in data analysis and problem-solving compared to GPT-4o.

• Reasoning Capabilities: Secured top positions in multimodal deep reasoning across China and benchmarked against leading international standards.

• Global Memory: First in China to achieve long-form video understanding, supporting in-depth reasoning and analysis for videos up to 10 minutes long.

SenseTime remains committed to delivering smarter and more convenient AI experiences through continuous technological innovation. In May 2024, SenseTime launched the Sensechat Multimodal Large Model (Cantonese), the industry’s first Cantonese large model that supports colloquial Cantonese. It was subsequently made available to all Hong Kong residents in July of the same year for free. This latest upgrade to Sensechat marks another significant step forward in the localization of AI applications in Hong Kong. Looking ahead, SenseTime will continue to prioritize user needs and drive the widespread adoption of AI, uplifting our communities through technology.

Scan the QR codes below to download or update the Sensechat app. The app will be progressively rolled out on Google, Xiaomi, and Huawei’s app stores.