Elorian, a Palo Alto startup valued at $300 million after securing $55 million in funding prior to launch, has introduced its inaugural model, named Glimpse, as reported exclusively by TechCrunch.
Currently available as a research preview with API access, Glimpse is launching alongside a waitlist-only chat interface, meaning the current offering consists primarily of a results table and a core thesis. The central thesis posits that while today's AI models excel at language and code generation, they still lack genuine visual perception.
Unlike most multimodal models that convert an image into tokens and then reason about it in text, Glimpse continuously revisits the source image. During the reasoning process, the model can crop specific regions, extract frames from videos, or execute code to analyze and compare visual data—a mechanism Elorian refers to as a "visual scratch space." This functionality mimics a user pinching to zoom on a phone screen to verify details.
Elorian CEO Andrew Dai, a former Google DeepMind researcher, highlighted the limitations of current AI during a board game night. After photographing a game and asking a frontier model who was winning, the AI failed to answer. Dai explained that board games utilize unique visual languages and iconography that are intuitive for humans but remain difficult for AI models to interpret.
"Humans can look at IKEA diagrams, with no text, and assemble IKEA furniture, right? Why can't AI models?" Dai remarked. While this approach is not entirely new—OpenAI introduced similar crop-and-zoom capabilities to its o-series models in 2025—Elorian differentiates itself by building the entire model architecture around this feature at a lower cost. The company claims a small team trained Glimpse on a few hundred GPUs, resulting in a model significantly smaller than the frontier systems it is benchmarked against.
Regarding pricing, Glimpse costs $0.60 per million input tokens and $3 per million output tokens. This places it between mid-level models like Sonnet and Sol and lightweight models like Haiku and Luna.
Despite the pricing, Elorian's internal evaluations suggest Glimpse outperforms even GPT-6 Astra on tasks involving scientific chart reasoning and short-form video captioning, while maintaining competitive results across other visual reasoning benchmarks.
Elorian emerged from stealth six months ago following a seed round co-led by Striker Venture Partners, Menlo Ventures, and Altimeter Capital, with participation from Nvidia and Google chief scientist Jeff Dean.
The initial market focus is on companies with extensive video data, such as robotics firms, scientific laboratories, and media or creative enterprises. Looking toward the long term, Elorian aims to transition from simple visual perception to action, utilizing scene models for robotic control. Dai stated that by next year, the company intends to establish "the world's best visual reasoning model," aiming to lead on every available visual reasoning test.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.