{"id":648419,"date":"2026-10-16T17:30:09","date_gmt":"2026-10-16T17:30:09","guid":{"rendered":"https:\/\/www.newsbeep.com\/il\/648419\/"},"modified":"2026-10-16T17:30:09","modified_gmt":"2026-10-16T17:30:09","slug":"elorians-first-ai-model-wants-to-reason-with-images-not-just-about-them","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/il\/648419\/","title":{"rendered":"Elorian\u2019s first AI model wants to reason with images, not just about them"},"content":{"rendered":"

Elorian<\/a>, the Palo Alto AI startup that raised a $55 million seed round at a $300 million valuation before it launched a product, released its first model on Friday, TechCrunch has exclusively learned.<\/p>\n

The model, Glimpse<\/a>, is launching as a research preview with API access and a chat interface is waitlist-only, so for now it\u2019s a results table and a thesis. The thesis? Today\u2019s AI models, for all their skill with language and code, still can\u2019t really see.\u00a0<\/p>\n

Most multimodal models look at an image once, convert it to tokens, and then reason about it in text \u2014 like describing a photo from memory. Glimpse keeps going back to the source. Mid-reasoning, it can crop into a region, pull specific frames from a video, or run code to measure or compare what it\u2019s looking at, a mechanism Elorian calls a \u201cvisual scratch space.\u201d Think of someone pinching to zoom on a phone photo to double-check what they thought they saw.<\/p>\n

For former Google DeepMind researcher and Elorian CEO Andrew Dai, one of the clearest signs that AI still couldn\u2019t really see came on game night.\u00a0<\/p>\n

Last year he photographed a board game he\u2019d just finished with friends and asked a frontier model who was winning. It couldn\u2019t say. Board games \u201ceach have their own visual language, each have their own iconography, symbols,\u201d he told TechCrunch. \u201cTo an adult, it\u2019s quite easy to understand what\u2019s going on with minimal text, but these models aren\u2019t able to intuit what the symbols mean.\u201d<\/p>\n

\u201cHumans can look at Ikea diagrams, with no text, and assemble Ikea furniture, right? Why can\u2019t AI models?\u201d he said.\u00a0<\/p>\n

The approach isn\u2019t completely new \u2014 OpenAI added crop-and-zoom during reasoning to its o-series models in 2025 \u2014 but Elorian\u2019s pitch is that it built the whole model around it, cheaply. The company says a small team trained Glimpse on a few hundred GPUs and that the model is substantially smaller than the frontier systems it\u2019s benchmarked against.\u00a0<\/p>\n

Glimpse costs $0.60 per million tokens in and $3 per million tokens out \u2014 cheaper than mid-level models like Sonnet and Sol but more expensive than lightweight models like Haiku and Luna.<\/p>\n

But according to Elorian\u2019s evaluations, Glimpse outperforms even GPT-6 Astra on scientific chart reasoning and short-form video captioning, with competitive results on other visual reasoning tasks.\u00a0<\/p>\n

Elorian came out of stealth six months ago with a round co-led by Striker Venture Partners, Menlo Ventures, and Altimeter Capital, with participation from Nvidia and Google chief scientist Jeff Dean.<\/p>\n

The initial targets are companies sitting on lots of video: robotics firms, scientific labs, media, and creative companies. Longer term, Elorian wants to move from seeing to doing. That means building a model of a scene using that to choose an action, including directly controlling robots. Next year, Dai said, the company aims to build \u201cthe world\u2019s best visual reasoning model,\u201d one that leads on every visual reasoning test.\u00a0<\/p>\n

When you purchase through links in our articles, we may earn a small commission<\/a>. This doesn\u2019t affect our editorial independence.<\/p>\n","protected":false},"excerpt":{"rendered":"Elorian, the Palo Alto AI startup that raised a $55 million seed round at a $300 million valuation…\n","protected":false},"author":2,"featured_media":648420,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[7928,6947,85,46,125],"class_list":["post-648419","post","type-post","status-publish","format-standard","has-post-thumbnail","category-technology","tag-ai-models","tag-exclusive","tag-il","tag-israel","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts\/648419","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/comments?post=648419"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts\/648419\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/media\/648420"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/media?parent=648419"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/categories?post=648419"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/tags?post=648419"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}