Elorian releases Glimpse, its first visual-reasoning model

Palo Alto startup Elorian has released Glimpse, its first AI model, as a research preview. TechCrunch reported that the model is built around revisiting source images during reasoning, an approach the company presents as central to its visual reasoning work.

Crypto Briefing’s headline also describes Glimpse as an AI model built to reason in pictures, while Traders Union’s headline positions it in the enterprise AI market. The model is available through an API, but Elorian’s chat interface is limited to users on a waitlist, according to TechCrunch. The outlet described this as the company’s first model release.

The article identifies Elorian as a Palo Alto AI startup; it does not provide further details about the product’s user base or the scope of the research preview. TechCrunch reported that Elorian’s approach centers on having the model return to image material as it reasons.

The article places that approach alongside an earlier technique: it says OpenAI added crop-and-zoom during reasoning to its o-series models in 2025. TechCrunch characterizes Elorian’s pitch as building the model around this capability, but the report does not describe the implementation in technical detail.

Elorian prices Glimpse at $0.60 per million input tokens and $3 per million output tokens, TechCrunch reported. The outlet says those rates are below mid-level models such as Sonnet and Sol, but above lightweight models such as Haiku and Luna. These comparisons are presented as pricing context; the article does not state the prices of those other models.

The performance claims in the article come from Elorian’s own evaluations. The company said Glimpse outperformed GPT-6 Astra on scientific chart reasoning and short-form video captioning, and delivered competitive results on other visual reasoning tasks.

TechCrunch does not describe independent testing of those claims, so the reported results should be understood as the company’s evaluation rather than an independently established ranking. Elorian’s stated longer-term goal is to build what it calls “the world’s best visual reasoning model” next year, TechCrunch reported.

The article gives no specific benchmark threshold for that ambition. For now, the reported release is a research preview with API access and a waitlist-only chat interface, alongside the company’s claims about pricing and evaluation performance.