Chandra is a multimodal model from datalab-to that accepts both text and image inputs and produces text outputs. Its open-weight nature means its internals are inspectable and locally deployable. Beyond these structural facts, its distinctive behaviors and trade-offs aren't well-documented in available sources.