Chinese AI labs stay quiet after GPT-6, with GLM-Omni and Kimi K3.1 traces appearing

X user X.PIN says Chinese models have been quiet for over a month since OpenAI released GPT-6 and Anthropic released Fable 5.5, and argues that distillation alone will not close the gap with those labs.

He points to a recent paper co-authored by Z.AI founder Tang Jie that mentions GLM-Omni, which he speculates is GLM-5.5, a native multimodal model. The paper reportedly shows task accuracy rising from 12.57% to 26.86% through interleaved image-text alignment, a figure the poster cites without a separate source. He also says Kimi K3.1 references have appeared in SGLang commits that keep the K3 tokenizer and chat format and add a "medium" reasoning tier. None of these details has been confirmed by the labs.