GitHub radar
ZDTaichu 5.0: local 9B model with spatial vision
TaichuAI has published ZDTaichu5.0-9B, a multimodal model with image understanding, video understanding, spatial awareness. It’s locally available in 4-bit mode and uses ~6 GB of memory.
ZDTaichu5.0-9B is “a multimodal foundation model for general image understanding, spatial reasoning, agentic tool use, and embodied AI research” according to the authors. It’s based on TaichuAI’s combination of the Qwen3.5-9B language model and the C-RADIOv4-H vision encoder. It supports inputs of text, images, videos. It has a 128 thousand token context length. It can perform spatial reasoning in an image, explaining where objects are and how they relate to each other. It has the best performance for spatial reasoning and agentic tasks among models with up to 10 billion parameters. The 4-bit version is locally available and can run on any laptop with ~6 GB of memory. Licensed under NVIDIA Open Model License Agreement.
Why a vibe-coder should care
This model could be used for tasks involving looking at an image and understanding its spatial context (explaining a room, reading diagrams, understanding user interfaces). It’s a good local choice for 9B models, and I would use it for basic repetitive tasks with images without paying for a cloud subscription. The model also has interesting performance on agentic tasks for its size. The main limitation is that it does not have an available Ollama file and requires some configuration via Python.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install the ZDTaichu5.0-9B multimodal model from TaichuAI for me: https://huggingface.co/TaichuAI/ZDTaichu5.0-9B — find the right variant for my machine, set it up with transformers, and show me how to ask it a question about an image.
A regular laptop with 16 GB of RAM or more — the 4-bit version takes about 6 GB.
Open on Hugging Face▌ More finds