Z.ai founders were one of the pioneers in MM-LLMs with Cog-VLM years ago, back when LLaVA emerged. I wouldn't be surprised if they added multi-modal capabilities