Elorian Launches Glimpse, a Visual Reasoning AI Model
Elorian, a Palo Alto-based startup that raised $55 million in seed funding at a $300 million valuation before launching any product, has released its first AI model called Glimpse. The research preview includes API access and a chat interface. Unlike most multimodal models that convert images into tokens for reasoning, Glimpse continuously interacts with the source image during processing—cropping regions or analyzing video frames in real time through what Elorian calls a 'visual scratch space.' This approach aims to improve AI's ability to interpret complex visual data like board games and diagrams. The model is smaller than leading systems but outperforms them on certain tasks, such as scientific chart reasoning. Glimpse costs $0.60 per million input tokens and $3 per million output tokens—more expensive than lightweight models but cheaper than mid-tier ones. Elorian plans to target industries with large video datasets like robotics, media, and science labs. The company also aims for long-term development toward AI that can not only see but act on visual information by controlling robots. Founded in stealth mode six months ago, the startup was co-led by Striker Venture Partners, Menlo Ventures, Altimeter Capital, with participation from Nvidia and Google chief scientist Jeff Dean.
Topics: Artificial Intelligence Multimodal AI Startups Visual Processing