Chinese technology company Baidu has introduced two new artificial intelligence (AI) models, Ernie 4.5 and Ernie X1, asserting that they surpass those of DeepSeek and OpenAI based on specific performance metrics.
Baidu’s latest moves come as the competition among large language models (LLMs) intensifies. Ernie 4.5, a multimodal foundational model, and Ernie X1, specializing in multimodal reasoning, were made freely available on the Ernie Bot website on Sunday.
According to a Baidu statement posted on WeChat, Ernie 4.5 demonstrates superior multimodal capabilities, encompassing images, audio, and video, outperforming OpenAI’s GPT-4o on several benchmark platforms, including CCBench and OCRBench. Furthermore, the foundational model’s text capabilities were reported to be better than DeepSeek V3 and to roughly match OpenAI’s GPT-4.5 on various benchmarks.
Baidu, which was the first major Chinese tech firm to launch an LLM in March 2023 in response to the popularity of OpenAI’s ChatGPT, has faced increasing competition from other Chinese AI developers in the past two years. Baidu’s recent announcement is an attempt to strengthen its position in China’s AI landscape, particularly in response to DeepSeek’s open-source initiatives. Companies like Alibaba, Tencent, and Bytedance have also made significant progress in gaining business and consumer users of their AI models.
While Baidu did not provide detailed benchmark results for Ernie X1, the company stated that it “delivers performance on par with DeepSeek R1 at only half the price.”
For businesses seeking access to Ernie X1’s Application Programming Interface (API), the pricing is set at 2 yuan (US$0.28) per million token inputs and 8 yuan per million token outputs.
