China’s artificial intelligence industry has taken another major step forward with two significant announcements from Alibaba and DeepSeek. While Alibaba has introduced its largest and most advanced AI model so far, DeepSeek has focused on making AI much cheaper to use. Together, these developments show that the race in artificial intelligence is no longer just about building smarter models, it is also about making them faster, more efficient, and more affordable.
Alibaba recently launched Qwen3.8-Max, the biggest AI model in the company’s history. The model contains 2.4 trillion parameters, making it one of the largest AI models announced to date. However, what makes it particularly efficient is its Mixture-of-Experts (MoE) architecture. Instead of activating the entire model for every request, it uses only the parts needed for a specific task. In practice, around 95 billion parameters are active during each query, helping reduce computing costs while maintaining strong performance.
The model is also designed to handle more than just text. Qwen3.8-Max supports multimodal capabilities, allowing it to understand and process text, images, and videos. Another notable feature is its one-million-token context window, which enables the model to work with extremely long documents, detailed reports, large codebases, and lengthy conversations without losing track of earlier information.
Alibaba has also released Qwen3.8-Max as an open-weight model. This means developers and businesses can access and customize the model for their own applications instead of relying solely on a closed API. Open-weight models have become increasingly popular because they provide greater flexibility for organizations building AI-powered products and services. Following the announcement, Alibaba’s shares listed in Hong Kong rose as investors responded positively to the company’s latest AI milestone.
On the same day, Chinese AI startup DeepSeek announced V4-Flash, a model that stands out for a different reason. Rather than focusing only on size, DeepSeek has prioritized affordability. According to AI benchmarking firm Artificial Analysis, V4-Flash currently ranks as the lowest-cost well-known AI model in its database.
The reported pricing is $0.14 per million input tokens and $0.28 per million output tokens, making it significantly less expensive to operate than many other leading AI models. Artificial Analysis also estimates that running one of its benchmark tests on V4-Flash costs only about $0.03, highlighting how quickly AI inference costs are falling.
Despite its lower operating cost, V4-Flash still delivers competitive performance. Artificial Analysis gives the model an Intelligence Index score of 50, placing it alongside several other high-performing fast AI models. DeepSeek has also confirmed that the model supports multimodal inputs and long-context processing, while a more advanced version, V4-Pro, is reportedly under development.
These launches reflect a broader shift taking place across China’s AI industry. Companies including Alibaba, DeepSeek, ByteDance, Tencent, Baidu, and Moonshot AI are investing heavily in developing advanced AI systems. At the same time, they are working to lower the cost of running these models, making AI more accessible to businesses, developers, researchers, and startups.
The growing popularity of open-weight AI models is another important trend. Unlike fully proprietary systems, open-weight models allow organizations to inspect, fine-tune, and deploy AI on their own infrastructure. This gives businesses greater control over how they use AI while encouraging wider adoption across industries.
The latest announcements also highlight how competition in artificial intelligence is evolving. For the past few years, much of the focus has been on building increasingly powerful models. Today, efficiency has become just as important. Companies are competing not only to deliver better performance but also to reduce the cost of using AI at scale.
Lower inference costs can make advanced AI tools more practical for organizations with limited budgets, opening the door for wider adoption across sectors such as education, healthcare, software development, finance, and customer service.
Together, Alibaba’s Qwen3.8-Max and DeepSeek’s V4-Flash represent two different approaches to the same goal: making artificial intelligence more capable and more accessible. As AI continues to evolve, performance and affordability are becoming equally important, shaping the next phase of global competition in the rapidly growing AI industry.
