Spatial AI lab building foundation models that understand the 3D world
Speridlabs is a spatial AI research lab building foundation models that perceive, reason about, simulate and generate 3D worlds, for uses such as robotics, gaming and visual effects. Its upcoming 3D world model, which TechCrunch reports is called Mundus, builds persistent, queryable 3D scenes and objects from a text prompt or a handful of images. The lab also publishes open research, including Iris-3B, a fully open pixel-space text-to-image model, and ENEAS, a text-promptable method for instance tracking and segmentation. Speridlabs was part of the PearX accelerator batch that presented at Pear VC's demo day in San Francisco in late September 2026.