Elverian's first AI model wants to reason with images, not just about them
According to Elorian's evaluations, Glimpse even outperforms the GPT-6 Astra in graph scientific reasoning and short video descriptions, with competitive results in other visual reasoning tasks.
Last day to save $200 on the TechCrunch Founders Summit on October 16th. Register now.
Disrupt the doors opening on October 13th. Get your card and bring someone with you with a 50% discount. Register now.
Elorian, the Palo Alto AI startup that raised $55 million at a $300 million valuation before launching a product, released its first prototype on Friday, TechCrunch has learned exclusively.
The Glimpse model is launched as a research preview with API access, and the chat interface is waitlisted only, so it's currently the results table and thesis. Thesis? Today's AI models, for all their language and code skills, still can't really see.
Most multimodal models look at an image once, convert it to a token, and then reason about it in text—like describing a photo from memory. Glimpse keeps returning to the source. On average, it can cut into an area, pull specific frames from a video, or run code to measure or compare what it's looking at, a mechanism Elverian calls "visual scratch space." Think of someone zooming in on a phone photo to check what they think they see.
For former Google DeepMind researcher and Elorian CEO Andrew Dye, one of the clearest signs that AI still couldn't really see happened on game night.
"Humans can look at IKEA diagrams without text and assemble IKEA furniture, right? Why can't AI models?" he said
This approach isn't entirely new—OpenAI added cropping and zooming during reasoning to its o-series models in 2025—but Elorian's step is to build the entire model around it, cheaply. The company says a small team has trained Glimpse on a few hundred GPUs, and the model is significantly smaller than frontier systems.
Glimpse costs $0.60 per million tokens and $3 per million output tokens – cheaper than mid-tier models like Sonnet and Sol but more expensive than lightweight models like Haiku and Luna.
But based on Elorian's evaluations, Glimpse outperforms even the GPT-6 Astra in graph scientific reasoning and describing short videos, with competitive results in other visual reasoning tasks.
Elverian came out of hiding six months ago with a round led by Striker Venture Partners, Menlo Ventures and Altimeter Capital, with participation from Nvidia and Google chief scientist Jeff Dean.
The primary targets are companies that sit on a lot of video: robotics companies, science labs, media and creative companies. In the long term, Elverian wants to move from seeing to doing. This means building a model of the scene using it to select an action, including directly controlling the robots. Next year, Dai said, the company plans to build "the world's best visual reasoning model," one that tops every visual reasoning test.
When you make a purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
Get 50% off second pass. Cover more ground at TechCrunch's ultimate founder bootcamp of the year with your partner. Co-founder or partner 50% off second pass. Meet scale and investment leaders, ask questions, find solutions and more. Last day to save October 16
Tesla unveils SpaceX Roadster that flies and crashes Sean O'Kin's style.
The FBI arrested an OpenAI employee accused of planting a secret recording device in Zack Whittaker's conference room
Bumble agrees to $1.3 billion acquisition deal as online dating faces Lauren Forrestal reset