On October 13, 2026, domestic large‑model vendor StepStar and Alibaba’s Qwen both unveiled AI‑focused consumer hardware on the same day: StepStar launched the STEPX Neo “Early Access” AI phone, priced at 3,999 yuan across the full stack, with the built-in personal agent Amoo; meanwhile, Alibaba introduced the Qwen AI glasses N1/N1 Pro and earclip‑style headphones, and revealed that its Wuying team is piloting the QwenBook AI tablet, with a full‑featured version expected by late 2026 or early 2027. This move is no isolated incident—it reflects a strategic shift in China’s large‑model industry, moving from competing on model parameters and pricing to capturing high‑frequency user touchpoints through hardware such as smartphones, smart glasses, and earphones.
This shift stems from severe user‑retention pressures. According to QuestMobile data, in January–February 2026, the 30‑day retention rate for mainstream domestic AI apps stood at just 12.8%; meanwhile, model‑iteration cycles have accelerated to an average of one flagship release every 9.5 days, and in the first half of the year, the industry’s six price cuts have collectively driven a cumulative decline exceeding the total of the previous two years. With virtually zero switching costs at the software layer, large‑model providers struggle to build a stable user base and a viable monetization pathway. Yin Qi, CEO of Jieyue Xingchen, stated bluntly: “AI that remains confined to the app layer will always be a tenant; only by developing its own hardware can AI become a native inhabitant.”
In addition to StepStar and Alibaba, ByteDance’s DouBao launched the DouBao Phone in September 2026 (with over 360,000 pre-orders on JD.com) and, in partnership with China Mobile, introduced the “Lingxi·Xinghuo Smart Box” series, covering AI-powered smart notebooks, voice recorders, and AIPCs. Meanwhile, Xiaomi has deeply integrated its self-developed MiMo‑V2.6‑Pro large model—ranked first globally among open-source models as of September 2026—into the Xiaomi 18 Fold/Pro and the AI-native smartphone currently undergoing closed beta testing, codenamed “Lobster.” Amid fierce competition among multiple players, hardware is no longer merely a delivery platform; it has become the critical battleground for building user habits and closing the ecosystem loop around large models. However, challenges such as fragmented resources, high investment with a steep failure rate, and hardware manufacturers’ reverse engineering of large models further cloud this “battle for the entry point,” leaving it highly uncertain.