BEIJING — Chinese companies are accelerating the rollout of new artificial intelligence models as competition with U.S.-based rivals OpenAI, Anthropic, and Google intensifies.
Just over a year ago, China-made DeepSeek made headlines globally by launching an AI chatbot that challenged OpenAI's ChatGPT with competitive usage fees and production costs. This move raised concerns about the effectiveness of U.S. technology restrictions on Chinese firms.
On Tuesday, the Beijing-based startup Moonshot AI introduced Kimi K2.5, claiming it possesses video-generation and agentic capabilities that surpass those of OpenAI, Anthropic, and Google's leading models. Agentic AI refers to systems capable of performing tasks independently on behalf of humans, aiming for minimal user interaction and increased efficiency.
This announcement comes a mere three months after the initial release of the Moonshot's K2 model.
In related news, hours earlier, ecommerce giant Alibaba unveiled its latest generative AI model, designed to create text, pictures, or video from user prompts. According to Alibaba, the Qwen3-Max-Thinking model excelled in a significant benchmark test, known as "Humanity's Last Exam," outperforming major U.S. counterparts.
Alibaba highlighted that the model autonomously selects the optimal AI tool for various tasks and leverages previous conversations for context, generating responses more effectively and cost-efficiently.
A week earlier, on January 19, Z.ai launched a free version of its recently introduced GLM 4.7 model. Just two days later, the company had to limit new user registrations for their AI coding tool due to high demand exceeding their computing capacity.