The model, Glimpse, is being released as a research preview with API access, while its chat interface remains waitlist-only. For now, it amounts to a results table and a thesis.

The thesis: despite their prowess with language and code, today's AI models still cannot truly see. Most multimodal systems examine an image once, translate it into tokens, and then reason about it in text—akin to describing a photograph from memory. Glimpse, by contrast, keeps returning to the source. During reasoning, it can crop into a specific region, extract particular frames from a video, or execute code to measure or compare what it is examining—a mechanism Elorian refers to as a "visual scratch space." Picture someone pinching to zoom on a phone photo to double-check what they thought they saw.

For Andrew Dai, a former Google DeepMind researcher and Elorian's CEO, one of the most telling signs that AI still couldn't genuinely see arrived on game night. Last year, he photographed a board game he had just finished playing with friends and asked a frontier model who was winning. It couldn't answer. "To an adult, it's quite easy to understand what's going on with minimal text, but these models aren't able to intuit what the symbols mean."

"Humans can look at IKEA diagrams, with no text, and assemble IKEA furniture, right. Why can't AI models," he said.

The approach itself isn't entirely novel—OpenAI introduced crop-and-zoom during reasoning to its o-series models in 2025—but Elorian's claim is that it constructed the entire model around this capability, and did so 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 is benchmarked against. Glimpse costs $0.60 per million tokens in and $3 per million tokens out—less expensive than mid-level models like Sonnet and Sol but pricier than lightweight models like Haiku and Luna. Yet 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.

Elorian emerged from 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.

The initial targets are companies sitting on large quantities 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 and using it to choose an action, including directly controlling robots. Next year, Dai said, the company aims to build "the world's best visual reasoning model," one that leads on every visual reasoning test.