A month of silence for Chinese models after GPT-6: GLM-Omni and Kimi K3.1 revealed
AI summary
After the releases of GPT-6 and Fable 5.5, Chinese models fell silent for over a month, with some arguing that China can no longer catch up to OpenAI and Anthropic through knowledge distillation. A paper co-authored by Tang Jie mentions the native multimodal model GLM-Omni (presumed to be GLM-5.5), which raises task accuracy from 12.57% to 26.86% through interleaved image-text alignment. A reference to Kimi K3.1 appears in an SGLang commit, retaining the K3 tokenizer and chat format while adding a new "medium" reasoning tier; attention is now focused on Qwen 4, expected in Q4, and Hunyuan 4.1, expected in 2027.
Body
The English body is being prepared. You can open the original source.