The news: Alibaba is escalating China’s AI race with the launch of its Qwen2.5 Max AI model, claiming it outperforms OpenAI’s GPT-4o, DeepSeek-V3 and Meta’s Llama-3.1-405B in benchmark tests.
The context: The release follows the meteoric rise of DeepSeek, a Chinese startup whose AI models have shaken the industry for their high-performance capabilities that rival top-tier AI models at a fraction of the development cost.
DeepSeek’s open-source pricing strategy has forced major Chinese tech firms, including Alibaba, Tencent, and Baidu, into steep price cuts and rapid AI upgrades.
Alibaba previously slashed its AI model prices by up to 97% in response to DeepSeek’s low-cost approach. Other companies, including Tencent and Baidu, have also been forced to cut prices and accelerate model development to stay competitive.
The AI price war is shaping China’s AI industry, pushing major players to prioritise efficiency and cost-effectiveness.
With the launch of Qwen2.5 Max during the Lunar New Year holiday, Alibaba is showing it intends to remain competitive amid rising pressure from emerging startups like DeepSeek.
Meanwhile, the White House said the National Security Council would review potential security risks linked to DeepSeek’s expansion. And OpenAI said it found evidence the Chinese AI startup used its proprietary models to train its own.
The American firm told the FT it blocked suspected DeepSeek accounts last year over concerns of distillation—using OpenAI’s proprietary model outputs to train a competing AI, which violates OpenAI’s terms of service.