Elorian has opened a research preview of Glimpse, a model that can revisit visual inputs and use tools to analyze images and video.
Elorian released Glimpse on October 16, 2026, as a research preview available through an API and chat interface. The Palo Alto startup designed the model to work directly with visual material while reasoning, rather than relying only on a text description of what it has seen.
Glimpse can revisit an image during a task, zoom in on selected regions, extract frames from video, and run code to measure or compare visual details. Elorian calls this capability a “visual scratch space.” The company says the model was trained by a small team using a few hundred GPUs and is substantially smaller than the frontier systems used in its comparisons.
Elorian’s evaluations report that Glimpse outperformed GPT-6 Astra on scientific chart reasoning and short-form video captioning. The company also says it produced competitive results on other visual reasoning tasks. These are company-reported findings; the article does not provide independent validation of the evaluations. Glimpse is priced at $0.60 per million input tokens and $3 per million output tokens.
The model’s initial intended users include robotics companies, scientific labs, media businesses, and creative firms that work with substantial amounts of video. Elorian’s longer-term ambition is to use visual understanding to help systems select actions, including controlling robots, but that is a future direction rather than a capability announced as available in this release.
Elorian raised a $55 million seed round at a $300 million valuation before launching a product. The round was co-led by Striker Venture Partners, Menlo Ventures, and Altimeter Capital, with participation from Nvidia and Google chief scientist Jeff Dean. CEO Andrew Dai said the company aims to build a leading visual reasoning model next year.