Glimpse arrived on October 16, 2026 as the first model from Palo Alto startup Elorian, released as a research preview with API access while its chat interface stays waitlist-only, according to TechCrunch's exclusive report that day. The company's pitch is blunt: today's models can read text and write code, but they still cannot really see.
Most multimodal models look at an image once, convert it into tokens, and then reason about it in words, much like describing a photo from memory. Elorian CEO Andrew Dai, a former Google DeepMind researcher, tells a game-night story: he photographed a board game he had just finished and asked a frontier model who was winning. It could not say. Board games each have their own visual language and iconography that an adult grasps at a glance, yet the model could not intuit the symbols. Humans can assemble Ikea furniture from diagrams with no text at all; models cannot.
Halfway through reasoning, it goes back to the source
Glimpse keeps what Elorian calls a visual scratch space. Mid-reasoning, the model can crop into a region, pull specific frames from a video, or run code to measure and compare what it is looking at, instead of guessing onward from a first impression. The approach is not entirely new, since OpenAI added crop-and-zoom during reasoning to its o-series models in 2025, but Elorian built the whole model around it, and cheaply: a small team trained Glimpse on a few hundred GPUs, and the model is substantially smaller than the frontier systems it benchmarks against.
Treat the scorecard and the price tag with care
Glimpse costs $0.60 per million input tokens and $3 per million output tokens, cheaper than mid-level models like Sonnet and Sol but pricier than lightweight ones like Haiku and Luna. In Elorian's own evaluations, Glimpse outperforms even GPT-6 Astra on scientific chart reasoning and short-form video captioning, with competitive results on other visual reasoning tasks. Every one of those numbers comes from the company itself, and with access limited to a research preview and a waitlist, outsiders cannot yet reproduce them. For now the scorecard is a claim to test, not a verdict.
The first customers are companies sitting on video
Elorian came out of stealth six months ago with a $55 million seed round at a $300 million valuation before launching any product, co-led by Striker Venture Partners, Menlo Ventures and Altimeter Capital, with participation from Nvidia and Google chief scientist Jeff Dean. Glimpse first targets companies holding large amounts of video: robotics firms, scientific labs, media and creative companies. Longer term, Elorian wants to move from seeing to doing, building a model of a scene and using it to choose actions, including directly controlling robots, and it aims to build the world's best visual reasoning model next year. That is an ambition, not a result. The contest in visual models is shifting from whether a system can recognise what is in an image to whether it knows to go back and look again when it is unsure.
