Alibaba Reveals Improved AI Model, Says It Beats Competitor DeepSeek-V3

Alibaba, a Chinese tech company, claimed to have outperformed the DeepSeek-V3 with its latest Qwen 2.5 artificial intelligence model.

The Qwen 2.5-Max’s odd release date—the first day of the Lunar New Year, when the majority of Chinese are off from work and spending time with their families—indicates how much pressure DeepSeek’s explosive growth over the last three weeks has put on both its domestic and international competitors.

Alibaba’s cloud unit released a statement on its official WeChat account stating that “Qwen 2.5-Max outperforms almost across the board GPT-4o, DeepSeek-V3, and Llama-3.1-405B,” alluding to the most cutting-edge open-source AI models from OpenAI and Meta.

Also Read: Alibaba Introduces Qwen 2.5 AI Models, Enhances Generative AI Capabilities

Silicon Valley has been rocked by the February 10 release of DeepSeek’s AI assistant, which is powered by the DeepSeek-V3 model, and the February 20 release of its R1 model. The Chinese startup’s allegedly low development and usage costs have caused investors to doubt the massive spending plans of top AI companies in the US.

However, DeepSeek’s domestic competitors are rushing to improve their own AI models as a result of its success.

ByteDance, the company that owns TikTok, issued an upgrade to its main AI model two days after DeepSeek-R1 was made public. The company claimed that the model fared better than Microsoft-backed OpenAI’s o1 in AIME, a benchmark test that gauges how effectively AI models comprehend and react to complex instructions.

This supported DeepSeek’s assertion that, on several performance metrics, their R1 model was on par with OpenAI’s o1.

When DeepSeek’s V3 model’s predecessor, DeepSeek-V2, was published in May of last year, it set off a pricing war for AI models in China.

Also Read: Alibaba Unveils Wonder Avenue: AI Shopping Experience at Paris 2024 Olympics

Alibaba’s cloud division announced price reductions of up to 97% on a variety of models due to DeepSeek-V2’s open-source nature and historically low cost of just 1 yuan ($0.14) for 1 million tokens, or units of data processed by the AI model.

Other Chinese tech firms followed suit, such as Tencent, the most valuable internet company in the nation, and Baidu, which launched China’s first ChatGPT-like app in March 2023.

In a rare interview with Chinese media site Waves in July, Liang Wenfeng, the mysterious creator of DeepSeek, stated that the company “did not care” about price wars and that its primary objective was to achieve artificial general intelligence or AGI.

AGI is defined by OpenAI as autonomous systems that outperform humans in the majority of economically significant tasks.

Young graduates and PhD students from prestigious Chinese universities make up the majority of DeepSeek’s workforce, which functions more like a research lab than the hundreds of thousands of workers employed by major Chinese internet businesses like Alibaba.

Also Read: DeepSeek surpasses ChatGPT to lead the US Apple App Store

Liang contrasted DeepSeek’s lean operations and flexible management style with the exorbitant costs and top-down structures of China’s leading tech companies, stating in his July interview that he thought they might not be well suited to the future of the AI business.