GoKawiilElorian launches Glimpse, a vision AI model that can zoom and crop while reasoning
Palo Alto startup Elorian released its first model, Glimpse, as a research preview with API access and a waitlisted chat interface. Unlike typical multimodal models that convert an image to text tokens once, Glimpse can re-examine images mid-reasoning—cropping regions, pulling video frames, or running code to measure what it sees. The company says a small team trained it on a few hundred GPUs, and it costs $0.60 per million input tokens and $3 per million output tokens.
GoKawiil's interpretation of the reporting above, not reported fact.
Elorian's approach suggests a cheaper, smaller-scale path to improving visual reasoning rather than relying purely on scaling frontier models, which could pressure larger AI labs to justify their costs if performance proves competitive. CEO Andrew Dai's framing—that models still can't truly 'see' visual language like board-game symbols or diagrams—positions Glimpse as addressing a gap that text-centric multimodal systems have struggled with. The launch follows a $55 million seed round at a $300 million valuation before any product existed, raising the stakes for Glimpse to prove the startup's thesis in practice.
- Glimpse lets AI re-examine images mid-reasoning via cropping, video frame extraction, and code execution, rather than converting images to text once.
- Elorian trained the model on a few hundred GPUs, claiming it's smaller than the frontier systems it's benchmarked against.
- The model launches as a research preview with API pricing of $0.60 per million input tokens and $3 per million output tokens, with chat access currently waitlist-only.
Source: techcrunch.com — Aditya Mehta, 2026-10-16
Published there as: “Elorian’s first AI model wants to reason with images, not just about them”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.