Alibaba just dropped Qwen 3.5 and benchmarked it against GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro. they did not mention DeepSeek once. that tells you everything about where the real competition is. DeepSeek dominated the Chinese AI conversation for a year. now Alibaba is trying to leapfrog them by targeting US models directly. the specs: 397B open-weight model. 1M token context. native multimodal for the first time. 60% cheaper than Qwen 3. 8x batch efficiency. visual agentic capabilities for autonomous app navigation. this dropped alongside ByteDance’s Doubao 2.0. DeepSeek’s next model is expected in days. every major Chinese lab shipped a flagship this month. Lunar New Year timing. they’re all racing to lock in users before the holiday. 2026 is the year the US vs China model gap effectively closed. and open-weight is the weapon