Elorian, a Palo Alto-based AI startup, released its first model, Glimpse, on Friday as a research preview with API access and a waitlist-only chat interface, according to techcrunch.com. The company had previously raised $55 million in a seed round at a $300 million valuation before launching any product.
Glimpse differs from existing multimodal AI models by continuously referring back to the original image during reasoning rather than converting it once into text tokens. This approach, called a "visual scratch space," allows the model to crop into regions, extract specific video frames, or run code to measure and compare visual elements. Elorian CEO Andrew Dai, formerly of Google DeepMind, explained that current models struggle to interpret visual languages and symbols, such as those in board games or IKEA diagrams, which humans understand intuitively.
The model's ability to reason directly with images addresses a key limitation in AI's visual understanding, where most models only describe images from memory. By enabling ongoing visual analysis, Glimpse aims to improve AI's comprehension of complex visual contexts. This capability could impact applications in fields requiring detailed image interpretation, such as gaming, design, and video analysis.
Elorian's Glimpse is currently available as a research preview with API access and a chat interface accessible via a waitlist, as noted by techcrunch.com. The company’s launch marks a notable step in advancing AI models that integrate continuous visual reasoning.