Tencent Holdings Ltd. released four open-source variants of its HunYuan large language model on 4 Aug., offering parameter sizes of 0.5 billion, 1.8 billion, 4 billion and 7 billion. The lightweight models handle up to 256,000 tokens and feature separate fast and slow reasoning modes, allowing them to run on a single consumer GPU and target use cases ranging from smart-home speakers to in-car assistants and personal computers. Hours later, Alibaba Group’s Qwen team introduced Qwen-Image, a 20-billion-parameter text-to-image model based on the MMDiT architecture. Alibaba says the system achieves state-of-the-art in-pixel text rendering in both English and Chinese, broadening its open-weight portfolio beyond language to multimodal generation. The rapid succession of releases underscores intensifying competition among Chinese technology giants to court developers with freely available model weights, a strategy that contrasts with U.S. rival OpenAI’s decision to delay open-sourcing its latest systems over security concerns.