On April 25, 2025, Baidu founder Robin Li made headlines at the Baidu AI Developer Conference by announcing the launch of two new AI models, the Wenxin Large Model 4.5 Turbo and Wenxin Large Model X1 Turbo. These models are positioned as significant competitors to the existing offerings from DeepSeek, a leading player in China's AI landscape. Li emphasized that the pricing of these new models is aggressively low, with the X1 Turbo priced at just 1 yuan per million tokens for input and 4 yuan for output, which is only 25% of the cost of DeepSeek-R1. In fact, the 4.5 Turbo model has dropped in price by a staggering 80%, now costing just 0.8 yuan per million tokens for input and 3.2 yuan for output—40% of DeepSeek-V3’s pricing.
This strategic pricing move is part of Baidu's broader initiative to capture market share in the rapidly evolving AI sector. According to Li, the goal is to make advanced AI capabilities accessible to developers and businesses, thereby fostering innovation and application development. "The essence of innovation is the lowering of cost," Li stated during his keynote, underscoring that reduced costs would enable developers to focus on creating "more super interesting and useful" AI applications rather than being bogged down by high foundational model expenses.
Li's remarks come amid a backdrop of intense competition in the AI market. DeepSeek, which gained prominence with its R1 model earlier this year, is now facing criticism from Baidu. Li pointed out that demand for DeepSeek's text-only generative AI models is "shrinking," citing their limitations in handling multimodal content—such as audio, images, and video—while also suffering from higher rates of so-called "hallucinations," which refer to the generation of incorrect or nonsensical outputs.
In a clear swipe at DeepSeek, Li articulated that while the company has made significant strides, its models still exhibit "pain points" that Baidu aims to address with its new offerings. He noted that DeepSeek's focus on text-only processing fails to meet the diverse needs of customers who require more integrated solutions.
To further bolster its position, Baidu unveiled several new applications at the conference, including a highly persuasive digital human platform and a general multi-agent collaboration app called "Xinxiang." These tools are designed to enhance user interaction and facilitate more complex AI applications. Additionally, Baidu introduced the content operating system "Cangzhou OS," which aims to provide a comprehensive framework for developers working with AI technologies.
Li's comments reflect a broader trend within the industry, as companies scramble to enhance their offerings in response to DeepSeek's rapid rise. Baidu's new models are explicitly designed to outperform DeepSeek's more narrowly focused approach by incorporating multimodal capabilities. This shift not only positions Baidu as a strong player in the AI field but also signals a potential recalibration of the competitive landscape.
As part of its ambitious AI strategy, Baidu has also developed a cutting-edge computing cluster, composed of 30,000 Kunlun P800 AI chips. This infrastructure is intended to support the training of multiple large models with hundreds of billions of parameters, allowing for extensive fine-tuning and customization for various applications. Li emphasized that the second half of the AI market will increasingly hinge on price and application, predicting a shift from high-cost, high-performance models to more accessible solutions that can drive widespread adoption.
In recent months, the AI landscape has seen a flurry of activity, with competitors like Alibaba and Tencent also ramping up their efforts. Alibaba has released its open-source QwQ model, claiming it matches DeepSeek's performance with significantly less training data. Meanwhile, ByteDance and Tencent have launched their own AI assistants, Doubao and Yuanbao, respectively, further intensifying the competition.
Despite Baidu's criticisms of DeepSeek, the latter continues to maintain a strong presence in the market, focusing on future developments with its upcoming R2 and V4 models. The race to dominate the AI space is not just about technology but also about strategic positioning and the ability to meet evolving consumer demands.
As the AI sector continues to grow, Baidu's latest moves could accelerate the adoption of AI technologies across various industries. By lowering the entry barriers for developers and enhancing the capabilities of its models, Baidu aims to fuel competition among leading cloud and AI providers in China. The Chinese government has also shown increased support for the AI sector, recognizing its importance as the country seeks to assert itself in the global technological landscape.
Investors are closely watching Baidu's stock performance, which has seen a positive response following the announcement of these new models. The consensus rating for Baidu (BIDU) remains a moderate buy, with analysts suggesting a potential upside of over 23% based on current price targets. As the competition heats up, the coming months will be critical for Baidu as it seeks to solidify its position as a leader in the AI market.
