ELORIAN
Building the foundation of visual reasoning.
At Elorian, we believe building systems that natively understand and reason through the visual medium the way humans do is the most important work in AI today. We're enabling models to think through spatial relationships, physical constraints, design intent, and abstractions to reach complex, creative solutions that work in the physical world.
Visual reasoning precedes language. Babies first learn to understand and navigate the world around them through sight and touch long before they speak. A great coach can identify a flaw in an athlete's stride, swing, or stroke at a single glance. Designers convey ideas not through words, but through sketches and images. Scientists peer through lenses to observe the natural world at its grandest and tiniest scales. If AI is going to accelerate technological progress in these areas, we will need tools that, like humans, are grounded in visual reasoning from the very beginning.
Current AI systems remain dependent on text. Today's vision language models reason in a two-step process: first translating visual inputs into language, and then performing text-based reasoning, sometimes with tools. No matter how advanced the subsequent reasoning, this chain is fragile, limited, and prone to hallucination. Since many visual tasks cannot be precisely (or concisely) described, today's models struggle with tasks that even children can handle effortlessly.
While there has been impressive progress in models that can produce photos and videos, this generative capability does not directly translate to reasoning about visual content. The widening gap between how models perform on tasks like coding, and visual tasks where the information cannot be easily described in words, runs deeper than existing benchmark performance. Developing models capable of understanding the spatial, structural, and relational complexity of the visual world will require fundamentally new techniques.
... combines multimodal training with new architectures for multimodal reasoning. Rather than treating images as static inputs, we train models to directly interact with and manipulate visual representations: interpreting structure, relationships, and constraints. Over time, this produces systems that can move from simple perception toward higher-level reasoning.
These models have the potential to dramatically accelerate technology across a broad set of industries just as existing models have transformed cybersecurity. Engineers can refine and iterate a design to produce a lighter engine or quieter wing. In robotics, our models will form the backbone of more capable systems that can operate across new environments. In medicine and the sciences, they will empower researchers to better understand disease. They will turn satellite imagery into usable insights across weather monitoring, disaster response, and precision agriculture. We take seriously the responsibility that comes with this vast potential impact and are committed to building robust safeguards.
Elorian is founded by researchers who have spent their careers working on the core technologies behind today's AI systems, and have led breakthroughs across pretraining, data, and vision modeling. We are a small, focused team building a research and product lab pushing the next leap forward in reasoning.
Elorian is backed by $55 million in funding from Striker Ventures, Menlo Ventures, and Altimeter with participation from 49 Palms and others with prominent AI researchers including Jeff Dean. We will invest deeply in the research required to push this new frontier.
If AI is to become truly general, it must be able to understand not just language, but the structure and composition of the physical world we inhabit.
We believe visual reasoning is a critical step toward that future — if you do too, join us.
