From：Internet Info Agency 2026-08-11 18:03:00
SenseTime today announced the launch of SenseNova 6.8 FlashLite Preview, a multimodal intelligent agent model, along with the introduction of the SenseNova Token Plan. The model supports autonomous planning, continuous execution, multi-agent collaboration, and dynamic error correction, enabling it to autonomously complete end-to-end tasks once a user-specified goal is provided. It excels in long-horizon, complex tasks with three key capabilities: executing tasks spanning hundreds of steps across multiple phases over several hours; orchestrating more than ten specialized sub-agents in parallel under a central main agent; and performing integrated planning and reasoning across multimodal inputs—such as text and images—within a unified task trajectory. The model can be used for tasks like generating dynamic PowerPoint presentations, understanding multimodal content on web pages, and directly operating browsers and applications to accomplish office-related workflows. SenseNova 6.8 FlashLite Preview is now available for public preview, and users can access further details via the official SenseNova Token Plan website and its GitHub repository.
