Elorian launches Glimpse, an AI model built to reason in pictures

Palo Alto startup Elorian has launched Glimpse, an AI model designed specifically for visual reasoning. The company says its own testing shows Glimpse outperforming GPT-6 Astra on two tasks: reasoning about scientific charts and captioning short-form video. That is a bold claim from a company that only came out of stealth in April 2026. It is also, for now, a test graded by the student who sat the exam. What Glimpse is claiming According to Elorian’s evaluations, Glimpse came out ahead of GPT-6 Astra on scientific chart reasoning and short-form video captioning. On other visual reasoning tasks, the company describes its results as competitive rather than dominant. The bigger idea behind Glimpse is architectural. Elorian builds native multimodal models, systems meant to work directly on images, video, and spatial or physical information without first converting any of it into text. Elorian’s central bet is that this kind of direct visual reasoning, rather than a text-mediated workaround, is what high-quality physical-world intelligence requires. Its target uses include robotics, engineering, medicine, and scientific analysis. A benchmark that exposes the gap Elorian has been vocal a...