Elorian Releases Glimpse, Its First Visual Reasoning Model, Following a $55 Million Seed Round

Palo Alto artificial intelligence startup Elorian released its first model, Glimpse, as a research preview on Friday. The launch follows a $55 million seed round that valued the company at $300 million earlier this year. Most multimodal models convert images to tokens once and reason about them in text. In contrast, Glimpse uses a visual scratch space, letting it crop regions, pull video frames, or run code mid-reasoning.

CEO Andrew Dai stated that current AI models struggle with distinct visual languages, such as board game symbols or text-free assembly diagrams. Elorian trained Glimpse using a few hundred graphics processing units, resulting in a system substantially smaller than competing frontier models. Pricing is set at $0.60 per million input tokens and $3 per million output tokens.

According to company evaluations, Glimpse outperforms GPT-6 Astra on scientific chart reasoning and short-form video captioning. Initial targets for the technology include robotics, scientific, media, and creative firms. Elorian plans to expand from visual recognition into physical action and aims to build the leading visual reasoning model next year.