China’s artificial intelligence sector witnessed a major competitive escalation as tech giant Alibaba unveiled its largest model to date, while rival startup DeepSeek introduced a new ultra-low-cost system. Alibaba launched its flagship Qwen3.8-Max, featuring 2.4 trillion parameters, designed to process complex enterprise tasks across an expansive one-million-token context window. Utilizing a mixture-of-experts architecture that activates only 95 billion parameters per individual request, Alibaba aims to deliver high-level performance while significantly lowering compute latency and operational expenses. Meanwhile, Hangzhou-based DeepSeek released its V4-Flash model, setting a new global benchmark for ultra-low-cost AI operations. According to research firm Artificial Analysis, V4-Flash charges $0.14 per million input tokens and $0.28 per million output tokens, costing an average of just 3 cents per benchmark test—making it over 100 times cheaper to run than competing frontier models such as Anthropic’s Claude Fable 5. The simultaneous releases highlight how Chinese AI developers are shifting from raw parameter expansion toward balancing high performance with aggressive cost-efficiency, targeting enterprise adoption both domestically and globally while challenging Western tech leaders.
Chinese Tech Giants Escalate AI Battle as Alibaba Launches Flagship Model and DeepSeek Slashes Costs
