Elorian, the Palo Alto SI startup that raised a $55 million seed round at a $300 million valuation before launching a product, released its first model on Friday. The release marks the company’s entry into the vision-focused SI market with a model designed to interact with images dynamically rather than merely describing them.
What Happened
The model, Glimpse, is launching as a research preview with API access and a waitlist-only chat interface. Elorian’s thesis is that current SI models, despite their proficiency with language and code, lack true visual comprehension. Most multimodal SI systems process an image once, converting it to tokens for text-based reasoning. In contrast, Glimpse utilizes a mechanism Elorian calls a "visual scratch space," allowing the model to crop into regions, pull specific video frames, or run code to measure and compare visual data during the reasoning process.
CEO Andrew Dai, a former Google DeepMind researcher, illustrated the gap in current SI capabilities by noting that frontier models could not determine who was winning a board game from a photo due to their inability to intuit complex visual iconography. While OpenAI added crop-and-zoom capabilities to its o-series models in 2025, Elorian claims to have built Glimpse specifically around this architecture. The company reports that a small team trained Glimpse on a few hundred GPUs, resulting in a model substantially smaller than frontier SI systems.
Why It Matters
Glimpse is priced at $0.60 per million input tokens and $3 per million output tokens. This pricing places it above lightweight models like Haiku and Luna but below mid-tier options such as Sonnet and Sol. According to Elorian’s internal evaluations, Glimpse outperforms GPT-6 Astra in scientific chart reasoning and short-form video captioning, while delivering competitive results in other visual reasoning tasks. The company targets industries with high volumes of video data, including robotics, scientific labs, and media. Long-term, Elorian aims to transition from visual reasoning to action, using scene models to control robots directly. Dai stated the company intends to build "the world’s best visual reasoning model" by next year.
The Bottom Line
Elorian has entered the SI industry with Glimpse, a vision model that prioritizes iterative visual analysis over static image processing. The startup, backed by investors including Nvidia and Google chief scientist Jeff Dean, positions Glimpse as a cost-effective alternative for specific visual reasoning benchmarks, with future ambitions extending to robotic control.