SenseNova 6.8 Flash

SenseNova 6.8 Flash is SenseTime's multimodal agent model officially released on November 2–3, 2026, aimed at finishing real office and design jobs end-to-end rather than only generating snippets. Distinct from radar entries SenseNova 6.8 Code2Video and SenseNova-Loop-9B, the Flash GA build emphasizes office collaboration and design workflows: long-term memory, cross-session handoffs, and deliverables such as dynamic PPT, posters, websites, SVG and Blender 3D, with Figma integration.

It is available via SenseNova Studio / Raccoon, the SenseNova Token Plan, and related Skills repos. Closed commercial model via SenseTime's platform.

Why this name

• Official product name: SenseTime's blog, TechNode, TMTPost and @SenseTime_AI consistently use “SenseNova 6.8 Flash”.
• Focus: vendor claims strong results on agent collaboration, data analysis and presentation generation; marketing highlights proactive reminders, cross-stage memory, and Agentic 3D / Vibe Dynamics design.
• Access: Studio / Raccoon experience, Token Plan docs, OpenSenseNova/SenseNova-Skills.

Caveats

Parameter count, context window and post-promo pricing are still thinly documented; benchmarks are largely vendor/media-reported. Do not conflate with Code2Video or Loop variants in the same SenseNova 6.8 family.

Who it is for

• Teams that want cross-session office agents and dynamic presentation/design delivery.
• Existing Raccoon / SenseNova platform users evaluating the Flash GA build.

Getting started

1. Read SenseTime's release post and TechNode coverage for capability boundaries.
2. Try Design features in SenseNova Studio / Raccoon.
3. Wire the SenseNova Token Plan API; check SenseNova-Skills for agent skills.
4. Regression-test on your own long-running projects and Figma/deck pipelines—not demos alone.