SenseTime's new SenseNova 6.8 Flash Lite preview version is online: focusing on lightweight and agile, aiming at "delegated intelligence"

IT House reported on August 11 that SenseTime Technology announced today (11th) that today, the preview version of SenseNova 6.8 Flash Lite Preview, SenseNova’s new multi-modal agent model, is officially launched on the SenseNova Token Plan. This model focuses on being lightweight and agile, with autonomous planning, continuous execution, multi-agent collaboration, and dynamic correction capabilities. It can achieve complete end-to-end result delivery and evolve towards Delegated Intelligence: users only need to give goals, and AI will autonomously go through the entire process, while humans focus on result acceptance. According to IT House, the official version of the 6.8 Flash Lite model and the 6.8 Flash model with further improved performance will also be released in the near future. According to the introduction, focusing on the delivery requirements of long-term and complex tasks, SenseNova 6.8 Flash Lite Preview has achieved breakthrough enhancements in three key capabilities: Long text and long-term stability: It supports maintaining goals, constraints and key facts in long-term tasks that last hundreds of steps, across stages and lasts for several hours, and has the ability to autonomously roll back and re-plan in the event of failure. Sub-agent dynamic collaboration: The main Agent can dynamically schedule and organize more than ten professional Agents to complete information retrieval, data analysis, mathematical calculations, visual understanding and fact verification in parallel. Native multi-modal fusion: realizes fusion planning, reasoning, tool invocation and result verification of text, pictures, charts, documents, videos and application interfaces under the same task trajectory. This model can produce dynamic PPT, complete content writing, chart production and layout design, and supports native HTML dynamic presentation, claiming to break the limitations of traditional static typesetting; it can natively understand multi-modal information, application interfaces and local files on web pages, directly operate browsers and applications for you, and independently complete various office tasks, script writing and automated operations. The model is now officially open for experience at the following address: