Elorian unveiled its first model, Glimpse, on 2026‑10‑16. The model is offered as a research preview with API access and a wait‑list chat interface. Unlike most multimodal AIs that turn an image into static tokens, Glimpse can repeatedly look at the source, crop regions, pull video frames or run code – a feature the company calls a “visual scratch space.” The startup says the model was trained on a few hundred GPUs and is smaller than competing systems.
Why it matters
Video‑heavy firms can add low‑cost, interactive visual reasoning to their pipelines without building large‑scale models themselves.
