Elorian releases Glimpse, its first AI model for image reasoning

In short: Elorian has released Glimpse, its first AI model, which is designed to analyze images and video by repeatedly zooming in and checking details while it works.

What happened

Elorian, a Palo Alto startup, released its first AI model on Friday, according to TechCrunch. The model is called Glimpse, and it is available as a research preview.

Developers can access Glimpse through an API (a paid doorway that lets software talk to the model). Elorian also has a chat interface, but it is waitlist-only for now.

Elorian says many AI systems look at an image once, turn it into text-like data, and then answer questions from that snapshot, like describing a photo from memory. Glimpse is built to keep looking back at the original image or video while it reasons. It can crop into a specific area, pull frames from a video, or run code to measure and compare what it sees, which Elorian calls a “visual scratch space” (like pinching to zoom on your phone to double-check a small detail).

The company says Glimpse costs $0.60 per million tokens in and $3 per million tokens out. Tokens are small chunks of text and image data used for billing, similar to being charged by word count.

Elorian claims that, in its own tests, Glimpse beat GPT-6 Astra on scientific chart reasoning and short-form video captioning, and performed competitively on other visual reasoning tasks.

Why it matters

More work is now stored as images and video, not just documents, including lab results, factory footage, and robotics camera feeds. If an AI system can reliably read a chart or follow visual steps, it could help people check details faster and reduce mistakes, but independent testing will matter since the performance claims come from the company.

Source: TechCrunch AI