#Elorian
2 related articles
·2 min
Can Frontier Models Be Fooled by Visual Decoys? Elorian Breaks Through with Thinking in Images
Elorian releases the decoy counting benchmark, revealing shortcomings in frontier multimodal models' visual reasoning, and proposes Glimpse to let models think with images rather than text.
Read more →
·2 min
Elorian Glimpse: A Vision Model That "Looks Back" and Zooms In on Image Details During Reasoning
Elorian releases Glimpse, its first vision model that re-examines images, crops regions, and runs code mid-reasoning using a visual scratch space architecture. We break down the tech, $0.60/M token pricing, few-hundred-GPU training claims, use cases, and risks.
Read more →