New AI 'Glimpse' Can Actually See Images, Not Just Describe Them
This AI can zoom in and reason like a human—could change how we use photos and video.
Elorian, the Palo Alto AI startup that raised a $55 million seed round at a $300 million valuation before launching a product, released its first model, Glimpse, on Friday as a research preview with API access and a chat interface. Glimpse is built around a “visual scratch space”: mid-reasoning, it can crop into an image region, pull specific frames from a video, or run code to measure or compare what it’s looking at, rather than just converting an image to tokens once.
Elorian says a small team trained the substantially smaller model on a few hundred GPUs. Glimpse costs $0.60 per million input tokens and $3 per million output tokens.
According to Elorian’s evaluations, Glimpse outperforms even GPT-6 Astra on scientific chart reasoning and short-form video captioning, with competitive results on other visual reasoning tasks. Initial targets are companies with lots of video—robotics firms, scientific labs, media, and creative companies—and longer term Elorian wants to move from seeing to doing, including directly controlling robots; next year, CEO Andrew Dai said, the company aims to build “the world’s best visual reasoning model.”
- Glimpse can zoom in and re-examine images during reasoning, unlike most AI that just glances once.
- It's cheaper and smaller than top models but claims better performance on visual tasks like charts and video.
- Initial users will be companies with lots of video, but eventually it could help robots see and act.
Why It Matters
Better visual AI could soon help with everything from reading X-rays to assembling IKEA furniture.