Perceptron Mk1.5

Perceptron Mk1.5 is an embodied reasoning model that processes text, image, video, and audio to produce text responses with structured spatial annotations like points, boxes, polygons, and tracks.

Strengths

It handles multimodal input across video and audio in addition to text and images, and can ground responses with precise spatial or temporal annotations.

Best for

Tasks involving physical agents, scene understanding, object tracking, and spatial reasoning where structured output about locations or movements is needed.

Limitations

Specialized for embodied reasoning and spatial tasks; Mk1, the earlier general-purpose model, may be more suitable for text-centric workloads without the need for video, audio, or spatial annotations.