One year after DeepSeek, Chinese AI firms from Alibaba to Moonshot race to release new models

One year after DeepSeek's disruptive debut, Chinese artificial intelligence companies are accelerating the release of new AI models in a bid to keep pace with US rivals OpenAI, Anthropic, and Google. Beijing-based startup Moonshot AI recently unveiled Kimi K2.5, claiming video-generation and agentic abilities that surpass top US models. Just months earlier, it had released the K2 model. Alibaba followed with Qwen3-Max-Thinking, which it says outperforms major US competitors on the comprehensive 'Humanity's Last Exam' benchmark; the model can auto-select the best AI tool for tasks and leverage past conversations efficiently at minimal extra cost. Z.ai made its GLM-4.7 model free but soon restricted sign-ups for its AI coding tool due to soaring demand. Shares of Baidu surged to near three-year highs after it released Ernie 5.0, claiming superiority over Google's Gemini-2.5-Pro. Google DeepMind CEO Hassabis acknowledged China's AI may be only 'months' behind the US. A key differentiator for Chinese firms is their open-source, low-cost approach, which is driving adoption in emerging economies where customization and affordability matter. Microsoft reported that DeepSeek usage in Africa is two to four times higher than elsewhere. Morningstar analyst Ivan Su warns that excessive focus on benchmarks overlooks the real value of AI when integrated into existing ecosystems like Tencent's WeChat, which is using its Yuanbao chatbot to distribute 1 billion yuan in cash awards during the Lunar New Year festival, mirroring the 'red envelope' campaigns that propelled WeChat Pay's dominance.