Zidong Taichu open-sources ZDTaichu5.0-9B multimodal model with leading spatial embodied intelligence

A 9B-parameter multimodal model, ZDTaichu5.0-9B, is now open source from Zidong Taichu. The team says it ranked first on 8 of 9 international spatial understanding benchmarks and leads among 10B-class general models. It handles video object grounding, multi-view spatial reasoning, and embodied task planning while keeping image-text understanding, OCR, math reasoning, and coding in the top tier. A full-stack data pipeline and built-in adaptive loop reasoning are used to improve spatial embodied performance without sacrificing general ability. The licence was not stated.

Original sources (Chinese)