Baidu To Release New AI Model Ernie 5 In 2025
Baidu is gearing up to release the next version of its artificial intelligence model, Ernie 5, in the second half of 2025. The upgraded model will feature advanced multimodal capabilities, allowing it to process and convert between text, images, video, and audio according to a report from CNBC. This comes as competition in China's AI sector heats up, particularly from the startup DeepSeek, which has disrupted the market with its cost-efficient AI models.
Baidu’s Ernie model is already integrated across its services, from cloud storage to content creation tools. Its Wenku platform, which uses AI to generate presentations and documents, has seen significant growth, reaching 40 million paying users by the end of 2024. While Baidu has not officially announced Ernie 5’s launch, the company’s planned update is expected to boost its standing in the AI sector.
Baidu vs DeepSeek
Despite being one of the first Chinese companies to launch a ChatGPT rival, Baidu has struggled to gain widespread adoption for its Ernie series. While the company claims that their current model, Ernie 4, matches the capabilities of OpenAI’s GPT-4, competitors like ByteDance’s Doubao chatbot and DeepSeek’s models have gained more traction.
The growing rivalry between Baidu and DeepSeek highlights a shift in the AI landscape. While DeepSeek focuses on cost efficiency and reasoning capabilities, Baidu continues to emphasize computational power and infrastructure. Whether Ernie 5 can reclaim Baidu’s position as a leader in China’s AI race remains to be seen, especially as competitors like Alibaba and ByteDance continue to advance their own AI technologies.
DeepSeek made headlines in January with its reasoning-focused AI model, which performed at a level comparable to OpenAI’s technology while being significantly cheaper to train. The model’s low-cost success triggered a wave of concern among tech investors, leading to a worldwide stock sell-off. The shift in AI development towards efficiency rather than sheer computational power has put pressure on players like Baidu to optimize their investments. Baidu CEO Robin Li acknowledged this challenge, stating, “You just don't know when and where innovations come from.”
Baidu CEO says more needs to be spent on infrastructure
Li remains adamant that substantial spending on cloud infrastructure and data centers is still necessary for AI advancement. Speaking at the World Government Summit in Dubai, he argued, “The investment in cloud infrastructure is still very much required. In order to come up with models that are smarter than everyone else, you have to use more compute.” His remarks contrast with DeepSeek’s approach, which has shown that innovation and efficiency can rival high-budget AI training.
The broader AI industry has seen massive financial commitments from major U.S. tech firms, with Microsoft, Alphabet, Amazon, and Meta collectively spending $246 billion on tech in 2024. This figure is expected to rise to $320 billion in 2025 as companies race to develop superior models and expand their computational capacity. In comparison, DeepSeek’s V3 model was reportedly trained on just $6 million, challenging the notion that only deep-pocketed firms can compete in AI.
For now, Baidu is betting on multimodal capabilities and large-scale investment to differentiate itself. But with AI innovation becoming more unpredictable, the company may need to rethink its strategy to stay ahead in an increasingly competitive market.
