Verdict
"Glimpse is a results table and a thesis for now. The thesis, that AI still can't really see, is right, and that's what makes it worth watching."
GEO HIGHLIGHTS
- Palo Alto startup Elorian released its first model, Glimpse, on October 16, 2026, as a research preview with API access.
- Glimpse can crop regions, pull video frames or run code to measure an image while it reasons, which Elorian calls a "visual scratch space".
- Pricing is $0.60 per million input tokens and $3 per million output tokens.
- Elorian says Glimpse beats GPT-6 Astra on scientific chart reasoning and short-form video captioning, based on its own evaluations.
Most multimodal models look at an image once, turn it into tokens and then reason in text, a bit like describing a photo from memory. Glimpse keeps going back to the picture. It can zoom into a region, grab specific frames from a video or run code to compare what it sees.
Elorian came out of stealth six months ago after raising a $55 million seed round at a $300 million valuation, co-led by Striker Venture Partners, Menlo Ventures and Altimeter Capital, with Nvidia and Google chief scientist Jeff Dean also investing.
Elorian came out of stealth six months ago after raising a $55 million seed round at a $300 million valuation, co-led by Striker Venture Partners, Menlo Ventures and Altimeter Capital, with Nvidia and Google chief scientist Jeff Dean also investing.
Reality Check
CEO Andrew Dai, formerly at Google DeepMind, tells a simple story about why this matters. He photographed a finished board game and asked a frontier model who won. It couldn't say. "Humans can look at Ikea diagrams, with no text, and assemble Ikea furniture, right? Why can't AI models?" he told TechCrunch.The idea isn't brand new. OpenAI added crop-and-zoom during reasoning to its o-series models in 2025. Elorian's pitch is that it built the whole model around this loop, and did it cheaply: a small team, a few hundred GPUs and a model much smaller than the systems it benchmarks against. That shows in the price, which sits below mid-tier models but above lightweight ones.
The first customers Elorian is targeting are companies with lots of video: robotics firms, science labs, media and creative studios. The longer-term plan is to go from seeing to doing, including controlling robots directly.
💀 Critical Risks
- All benchmark results so far come from Elorian's own evaluations, not independent testing.
- The chat interface is waitlist-only; for now access is via API.
- Bigger labs already ship similar zoom-while-reasoning features and can move fast.
FAQ: Can I try Glimpse today?
Developers can use it through the API as a research preview. The chat interface is waitlist-only for now.
