Qwen 2.5 AI Model Outperforms DeepSeek-V3 and GPT-4o, Claims Alibaba

Chinese tech giant Alibaba has unveiled an upgraded version of its artificial intelligence model, Qwen 2.5, claiming that it outperforms top AI models like DeepSeek-V3, GPT-4o, and Llama-3.1-405B. The announcement was made on February 10, coinciding with the Lunar New Year, a time when most people in China are off work, which underscores the growing competition in the AI space.

Qwen 2.5 AI Model Outperforms DeepSeek-V3 and GPT-4o, Claims Alibaba

DeepSeek, a rising Chinese AI startup, has made waves in the industry with its DeepSeek-V3 model, launched on January 10, and its DeepSeek-R1, released on January 20. These releases have shaken up Silicon Valley and global AI markets due to DeepSeek’s low development and usage costs, forcing investors to rethink large-scale AI spending. DeepSeek’s success has not only pressured overseas competitors like OpenAI and Meta but has also intensified rivalry among Chinese AI firms, including Alibaba, ByteDance, Baidu, and Tencent.

Alibaba has introduced Qwen 2.5-Max, the most advanced version of its AI model, as a direct response to DeepSeek’s growing influence. According to Alibaba’s cloud computing unit, this new model delivers superior performance across various benchmarks, making it a formidable challenger in the AI race.

China’s AI Model Price War

DeepSeek launched DeepSeek-V2 in May 2023, starting the battle with China’s major tech firms. The model was not only open-source but also incredibly affordable, costing just 1 yuan ($0.14) per 1 million tokens. This pricing strategy disrupted the AI industry in China and prompted Alibaba to slash its cloud AI model prices by up to 97% in an attempt to stay competitive. Other major players, including Baidu and Tencent, also responded by adjusting their AI model pricing and capabilities.

Following DeepSeek’s R1 model launch on January 20, ByteDance, the parent company of TikTok, quickly updated its own flagship AI model. The company claimed that its new model outperformed OpenAI’s GPT-4o in AIME, a benchmark test that evaluates an AI’s ability to understand and respond to complex instructions. This claim echoed DeepSeek’s own assertion that its R1 model rivals OpenAI’s latest technology in multiple AI performance metrics.

DeepSeek’s AGI Ambitions and Lean Structure

Unlike major tech firms like Alibaba, Baidu, and Tencent, which employ hundreds of thousands of workers, DeepSeek operates more like a research lab. The startup consists primarily of young graduates and PhD students from China’s top universities, giving it an agile and cost-efficient structure.

DeepSeek’s founder, Liang Wenfeng, has expressed disregard for AI price wars, stating that the company’s ultimate goal is Artificial General Intelligence (AGI). OpenAI defines AGI as an advanced AI system that can outperform humans in most economically valuable tasks. In an interview in July 2023, Liang emphasized that large tech corporations are not necessarily well-suited for the AI industry’s future, citing their high costs and rigid structures as limitations compared to DeepSeek’s flexible and innovative approach.

See Also: Did DeepSeek Steal OpenAI’s Data? Microsoft and OpenAI Investigate

The Future of AI Competition in China

With Alibaba’s Qwen 2.5, DeepSeek’s V3 and R1 models, and ByteDance’s latest AI upgrade, China’s AI industry is experiencing rapid advancements. The intense competition is not only shaping the domestic AI landscape but also influencing the global AI race. As DeepSeek continues its AGI mission, and companies like Alibaba and ByteDance push for better AI performance, we will see who will emerge as the dominant player in China’s AI revolution.