Elorian company presented Glimpse, its first artificial intelligence model

Elorian has released Glimpse, its first visual reasoning model, featuring a visual scratch space that enables the system to crop image regions, extract video frames, and run code for analysis.

Built by a small team using a few hundred GPUs, the research preview is available with API access.

Glimpse stands out because it can process images and videos in a more flexible way than conventional models. During processing, it has the ability to focus on specific points in an image, isolate individual frames from a video, or run code to measure and compare visual data. The company calls this process a “visual workspace.”

According to Elorian CEO and former Google DeepMind researcher Andrew Dai, the model was trained by a small team using a limited number of computers. The cost of use was set at $0,60 per million input data units and $3 per million output units. Similar image-focusing technology was added by OpenAI to its own models in 2025.

According to Elorian’s tests, Glimpse outperformed previous models in understanding scientific diagrams and short video descriptions. The company had already raised $55 million in funding, reaching a valuation of $300 million before it even launched an official product. The funding included venture capital, Nvidia, and Google’s chief scientist Jeff Dean.

READ ALSO

MIT professor warns of the dangers of friendship with machines

The company's initial targets include robotics labs, science centers, and media and creative content companies. Elorian's next goal for the coming year is to create the world's best visual reasoning model.

G

Hacks.gr on Google

Set it as a preferred source for cybersecurity updates.

Preferred source

Read the original English article on Hacks.gr