According to reports, Palo Alto-based startup Elorian has launched its first visual reasoning model, Glimpse, currently available via API in research preview with chat interface access still on a waitlist. Glimpse enables cropping image regions, extracting specific video frames, and executing code for measurements or comparisons during inference—Elorian refers to this capability as the "visual draft space."

The company states that the model was trained by a small team using hundreds of GPUs, with a scale significantly smaller than its leading benchmark models; according to Elorian's internal evaluations, Glimpse outperforms GPT-6 Astra in scientific chart reasoning and short-video caption generation, while achieving competitive results across other visual reasoning tasks. Input pricing is $0.60 per million tokens, and output pricing is $3.00 per million tokens.