SenseTime Releases SenseNova 6.8 Flash, Supporting Long-Term Memory and Dynamic PPT and Poster Generation

On November 2, SenseTime launched the official version of SenseNova 6.8 Flash. Focusing on complex tasks in real-world environments, the model continues to explore the boundaries of intelligent capabilities and optimizes the key difficulties in task execution and result delivery. It not only supports cross-task and cross-session collaboration, but also extends its generation capabilities to richer delivery forms: from static web pages and dynamic content to design outcomes that combine aesthetic expression with practical value. In multiple benchmark evaluations, SenseNova 6.8 Flash surpasses some international frontier models in scenarios such as agent collaboration, data analysis, and presentation generation, placing it in a leading position among industry models. At the same time, SenseTime also examined the model's performance in scenarios such as office automation, information retrieval, multi-step collaboration, content generation, and long-context understanding. It is reported that SenseNova 6.8 Flash is open for experience starting today. From passive delegation to proactive interaction that better understands you, SenseNova 6.8 Flash is exploring a more proactive way of intelligent office interaction. Relying on long-term memory and proactive interaction capabilities, the model can continuously collaborate around users' goals and constraints, and when relevant information changes, identify matters that may be affected, propose adjustment suggestions, and help tasks continue to advance. Relevant benchmark evaluations further examined the model's factual recall, instruction following, conflict resolution, and memory of personal preferences and goals in long-term interactions. In a real office case, a user had recommended an enterprise AI product to a client, but the client declined cooperation because the product did not support private deployment. A month later, the product was upgraded and launched, exactly filling this capability gap. When the user was preparing materials for a new sales meeting report, SenseNova 6.8 Flash had already linked the new version's feature changes with the previously recorded client needs and proactively reminded you: this client who had once declined cooperation is now worth following up again. The model not only remembers why the client declined, but also further organizes a feature comparison to help you seize the opportunity for re-conversion in time. With the support of SenseNova 6.8 Flash, long-term memory prevents previously accumulated information from dissipating over time, allowing personal experience to settle into team knowledge; proactive interaction allows this accumulation to be reactivated when opportunities arise—every after-the-fact remedy can be transformed into an early warning. AI output must not only be "viewable" but also "usable." From a PPT, a poster, and a piece of dynamic content to the advancement of a task spanning several days, SenseNova 6.8 Flash hopes to make AI output not only "viewable" but also "usable." Dynamic PPT is not simply "adding animation" to an ordinary PPT. SenseNova 6.8 Flash incorporates "narrative rhythm" into its generation objectives: it first sorts out the information each page needs to convey, the places where pauses are needed, and the content that should be emphasized, and then arranges the motion effects accordingly. Transition types, element entrance order, chart growth methods, and layout specifications are considered in a unified way within the same generation, so that the motion effects in the final PPTX work in concert with the page structure. Dynamic posters, however, are not just about "making images move." SenseNova 6.8 Flash considers the main visual, text hierarchy, and motion arrangement together in the same generation: the movement range of the main visual coordinates with the position and timing of text appearance, and the timeline rhythm serves the information hierarchy. The generated results can be used for social media dissemination and can be extended into a series of assets according to the same visual language. In addition, SenseNova 6.8 Flash can also collaborate with Blender through an Agent to complete complex 3D creation. After the user inputs requirements, the Agent generates an initial version of the scene and continuously adjusts it based on actual rendering results and user feedback. For example, professional users can request the creation of a specific scene in Blender software (such as the Louvre), and then correct flaws in the first output, such as the Louvre roof and the position of the pyramid. Users can also add modification requests, such as adding weather animation effects. Through multiple rounds of "generation—observation—feedback—modification," the dynamic 3D scene is gradually improved, and the final output is a Blender project that can continue to be edited. In SVG design, the model supports vector graphics and animation creation, and can combine graphic construction, color matching, path control, and motion arrangement to generate vector designs with a consistent style that can be scaled losslessly and edited further. SenseTime stated that aesthetics and execution are two dimensions of the same result: design capability determines the presentation effect, while task understanding and sustained execution capability bring the results closer to real needs.