SenseNova 6.8 Flash

What it is

SenseNova 6.8 Flash is a multimodal agent model from SenseTime, whose official version the company announced in Hong Kong on 5 November 2026^[1]. Both words in that description carry weight. "Multimodal" means the model is built to work with more than text: its advertised outputs include layouts, graphics, web pages and 3D scenes. "Agent" means it is meant to carry a task through several steps rather than answer one question and stop. SenseTime frames it around office work and design production — settings where a request usually ends in a file that someone has to open, edit and pass along.

Memory and continuous work

The model adds enhanced long-term memory, proactive interaction and task execution, and it keeps continuity across conversations^[1]. In practice that combination points at work that spans more than one sitting. Memory means the context of a project can survive between sessions; continuity across conversations means a later request can build on an earlier one instead of starting over. Proactive interaction means the model is designed to move first sometimes — to suggest the next step or ask for what is missing — rather than wait to be told. Task execution means it runs the steps itself instead of only describing how they might be done.

What it can produce

SenseNova 6.8 Flash generates dynamic PPTX presentations, animated posters, web designs, SVG graphics and interactive Three.js 3D pages, and it builds 3D models and animation through Blender^[1]. It also works with design tools such as Figma^[1].

The output formats are the interesting part. A PPTX file is the format of ordinary presentation software, so the result arrives as editable slides rather than flat pictures. SVG is a vector format: SenseTime says the model's SVG output is losslessly scalable and further editable^[1], which means enlarging it does not blur the edges and its shapes stay separate objects that a designer can still change. Three.js is a JavaScript library for drawing 3D in a web browser, so pages built with it can be rotated and explored by whoever opens them. Blender is a 3D modelling and animation program, so producing work there means scene files that can be reopened and adjusted. Figma is a collaborative interface design tool, and connecting to it lets generated work land inside the files a design team already uses.

What SenseTime claims about performance

SenseTime said the model performs strongly on agent collaboration, data analysis and presentation generation, without publishing scores^[1]. It also said the model outperformed some leading international models in multiple benchmark evaluations, without naming the benchmarks^[1]. Benchmarks are standardised test sets used to compare models against one another; without the numbers or the names of the tests, the claim records the company's position but cannot be checked from the announcement alone.

How to try it

The model is available for free trial through the SenseNova Token Plan, with up to 1,500 free API requests every five hours^[1]. Developers can apply for an API key through the SenseNova Token Plan platform^[1], and the model supports cross-task collaboration^[1] — one task feeding into another inside the same working session rather than each request standing alone.

Where it sits

A related model, SenseNova 6.8-Flash-Lite, was among the models SenseTime named in its first-half 2026 results^[2], so the Flash name was already part of the company's product line before the full version was launched. With SenseNova 6.8 Flash, SenseTime is pushing a general model toward the specific work of making documents, graphics and 3D assets, and measuring itself against international models on that ground.