--- title: Elorian’s first AI model wants to reason with images,... description: Palo Alto startup Elorian, which raised $55 million in seed funding at a $300 million valuation before launching a product, released its first AI model,... canonical: https://daily.dev/posts/elorian-s-first-ai-model-wants-to-reason-with-images-not-just-about-them-lmvkmrop4 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Elorian’s first AI model wants to reason with images, not just about them | daily.dev og:description: Palo Alto startup Elorian, which raised $55 million in seed funding at a $300 million valuation before launching a product, released its first AI model,... og:url: https://daily.dev/posts/elorian-s-first-ai-model-wants-to-reason-with-images-not-just-about-them-lmvkmrop4 og:image: https://api.daily.dev/og/posts/LMVKmRop4.png og:image:alt: Elorian’s first AI model wants to reason with images, not just about them og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Elorian’s first AI model wants to reason with images, not just about them **[TechCrunch](https://daily.dev/sources/tc)** · 3 min read · 0 upvotes · 0 comments ## Summary Palo Alto startup Elorian, which raised $55 million in seed funding at a $300 million valuation before launching a product, released its first AI model, Glimpse, as a research preview. Unlike typical multimodal models that convert an image to tokens once and reason about it purely in text, Glimpse maintains a 'visual scratch space,' letting it crop into regions, pull specific video frames, or run code to re-examine visual content mid-reasoning. CEO Andrew Dai, a former Google DeepMind researcher, says this addresses failures like models being unable to read board-game states or IKEA diagrams. Glimpse costs $0.60 per million input tokens and $3 per million output tokens, positioning it between lightweight and mid-tier models, and Elorian claims it beats GPT-6 Astra on scientific chart reasoning and short video captioning despite being trained on just a few hundred GPUs. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much does Elorian's Glimpse model cost per token? Glimpse costs $0.60 per million input tokens and $3 per million output tokens. That places it above lightweight models like Haiku and Luna but below mid-level models like Sonnet and Sol. Elorian says Glimpse outperforms GPT-6 Astra on scientific chart reasoning and short-form video captioning despite being trained on only a few hundred GPUs by a small team. _Comparing vision-model pricing and benchmarks gets easier when tracked in one place on daily.dev._ ### What is a visual scratch space in an AI model? A visual scratch space is a mechanism, introduced by Elorian's Glimpse model, that lets the model return to the original image or video mid-reasoning instead of converting it to tokens just once. It can crop into a region, pull specific video frames, or run code to measure or compare what it sees, similar to zooming into a photo to verify a detail. _Developers tracking new multimodal reasoning techniques can follow updates like this on daily.dev._ ## Similar posts on daily.dev - [Be the Architect: Intentional AI Engagement for Cybersecurity Engineering](https://daily.dev/posts/be-the-architect-intentional-ai-engagement-for-cybersecurity-engineering-gda57eiyc) · SpecterOps · 0 upvotes · 0 comments - [Google's Gemini 3.5 Flash costs 3x the model it replaced, and the era of cheap AI is ending](https://daily.dev/posts/google-s-gemini-3-5-flash-costs-3x-the-model-it-replaced-and-the-era-of-cheap-ai-is-ending-ochkeykzb) · XDA Developers · 2 upvotes · 0 comments - [This startup thinks robotics is about to have its ChatGPT moment](https://daily.dev/posts/this-startup-thinks-robotics-is-about-to-have-its-chatgpt-moment-rovzaavdf) · TechCrunch · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#multimodal](https://daily.dev/tags/multimodal) [View this post on daily.dev](https://daily.dev/posts/elorian-s-first-ai-model-wants-to-reason-with-images-not-just-about-them-lmvkmrop4) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Elorian’s first AI model wants to reason with images, not just about them","url":"https://daily.dev/posts/elorian-s-first-ai-model-wants-to-reason-with-images-not-just-about-them-lmvkmrop4","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/elorian-s-first-ai-model-wants-to-reason-with-images-not-just-about-them-lmvkmrop4"},"datePublished":"2026-10-16T16:04:40.804Z","dateModified":"2026-10-17T14:32:48.534Z","description":"Palo Alto startup Elorian, which raised $55 million in seed funding at a $300 million valuation before launching a product, released its first AI model,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/461c1b5af2c7808970adf9de4e2e06ca?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/461c1b5af2c7808970adf9de4e2e06ca?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"TechCrunch","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"TechCrunch","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tc","url":"https://daily.dev/sources/tc"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/elorian-s-first-ai-model-wants-to-reason-with-images-not-just-about-them-lmvkmrop4","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,multimodal","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"TechCrunch","item":"https://daily.dev/sources/tc"},{"@type":"ListItem","position":3,"name":"Elorian’s first AI model wants to reason with images, not just about them"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/elorian-s-first-ai-model-wants-to-reason-with-images-not-just-about-them-lmvkmrop4#faq","mainEntity":[{"@type":"Question","name":"How much does Elorian's Glimpse model cost per token?","acceptedAnswer":{"@type":"Answer","text":"Glimpse costs $0.60 per million input tokens and $3 per million output tokens. That places it above lightweight models like Haiku and Luna but below mid-level models like Sonnet and Sol. Elorian says Glimpse outperforms GPT-6 Astra on scientific chart reasoning and short-form video captioning despite being trained on only a few hundred GPUs by a small team. Comparing vision-model pricing and benchmarks gets easier when tracked in one place on daily.dev."}},{"@type":"Question","name":"What is a visual scratch space in an AI model?","acceptedAnswer":{"@type":"Answer","text":"A visual scratch space is a mechanism, introduced by Elorian's Glimpse model, that lets the model return to the original image or video mid-reasoning instead of converting it to tokens just once. It can crop into a region, pull specific video frames, or run code to measure or compare what it sees, similar to zooming into a photo to verify a detail. Developers tracking new multimodal reasoning techniques can follow updates like this on daily.dev."}}]} ```