Baidu Opens Up to Open Source: Ernie 4.5 to Be Available to the Public | Llm training dataset free | Openai | How was chatgpt trained | Turtles AI

The AI ecosystem in China has undergone a major shift with the introduction of DeepSeek models, leading to more efficient use of computing infrastructure. Open source emerges as a strategic element for AI deployment, with Baidu poised to follow this direction.

Key points:

• DeepSeek has increased the use of AI clusters in China, previously underutilized at 20 percent.
• Demand for AI hardware is growing strongly, with a focus on advanced GPUs.
• Baidu announces Ernie 4.5 in open source, inspired by DeepSeek’s success.
• AI offsets Baidu’s advertising decline, despite China’s economic crisis.

AI in China is undergoing a profound evolution, with the debut of DeepSeek models radically transforming the use of computing infrastructure in the country. Prior to the introduction of these models, China’s AI clusters were only 20 percent exploited, signaling a large discrepancy between hardware capacity and available software. This gap made clear the importance of a mature software ecosystem to maximize the efficiency of available hardware. The adoption of DeepSeek R1 reversed this trend, leading to a surge in demand for computational resources and accelerating the use of advanced GPUs, including NVIDIA’s Hopper accelerators and high-end consumer GPUs. In parallel, the success of DeepSeek’s open source approach has prompted Baidu to make the code for its new Ernie 4.5 model available, with the goal of fostering greater AI deployment. Starting in April, Baidu will offer free premium chatbot services, while the open source release of Ernie 4.5 is scheduled for June. The decision was announced by CEO Robin Li, who acknowledged how the open source strategy can facilitate large-scale adoption, just as demonstrated by the DeepSeek experience. However, despite progress in the AI sector, Baidu posted a 2% decline in quarterly Q4 2024 revenues to 34.12 billion yuan, slightly above analysts’ forecasts. The 26 percent growth in the Cloud AI business helped mitigate a 7 percent decline in the online marketing division, which was affected by the downturn in China’s real estate market and a reduction in corporate advertising investment. In December, Baidu’s Ernie platform recorded 1.65 billion daily interactions, a sharp increase from 600 million in August, signaling growing interest in generative AI services. However, Baidu’s U.S.-listed shares fell 7 percent in morning trading, reflecting general economic uncertainties.

Against this rapidly changing backdrop, China is consolidating its leading role in the AI sector by adopting new strategies that combine computational power, software innovation and open deployment models.