Multimodal model

Seele02

Seele02 is a native multimodal foundation model on a Mixture-of-Transformers architecture, unifying text, 3D, and spatial signals in one representation. An understanding expert handles semantics, reasoning, and planning, while a generation expert handles geometry, texture, and consistency, fused by shared global self-attention. It reads a 3D scene, describes it in text, renders a new viewpoint, and edits spatial layout from natural language in one conversation. Tool use is native, for agents.