Elorian

Elorian develops multimodal foundation models focused on visual and spatial reasoning about the physical world. It plans to bring those capabilities to developers and enterprises through an API. As of September 2026, its models page described its first visual thinking foundation model as forthcoming later that year and offered an early-access waitlist.

Founded in 2025, Elorian’s co-founders include CEO Andrew Dai, Yinfei Yang and Seth Neel, whose backgrounds span language-model training, multimodal modeling and data-centric AI. The privately held company is based in Palo Alto, California, and has raised $55 million, with backing from Striker Ventures, Menlo Ventures and Altimeter, alongside NVIDIA, 49 Palms and AI researchers including Jeff Dean.

Its research includes visual chain-of-thought reasoning, in which intermediate steps operate on an image. In Dai’s example, a model counts red hotels by first marking candidate objects with bounding boxes, then narrowing the selection by color. Elorian develops specialized multimodal data and combines evaluations, agents, supervised fine-tuning and reinforcement learning to improve its models. Its intended applications include robot scene understanding, construction monitoring and design tools that extract geometric constraints for further validation.