Baidu unveils new AI models, Ernie 4.5 and Ernie X1, claiming they outperform DeepSeek and OpenAI’s models, as the company faces growing competition in China’s AI market.
 |
| Baidu introduces multimodal AI models Ernie 4.5 and Ernie X1, positioning them as superior to DeepSeek and OpenAI models while aiming to expand its AI presence in China. Image: CH |
Beijing, China — March 16, 2025:
Baidu, a leading force in China’s technology sector, has unveiled two groundbreaking AI models, positioning them as superior to those of DeepSeek and OpenAI, amid growing competition in the large language model (LLM) market, reads a SCMP report.
On Sunday, Baidu introduced Ernie 4.5, a multimodal foundational model, and Ernie X1, its first multimodal reasoning model. Both models are now freely available on Baidu's website. According to Baidu, Ernie 4.5 has outperformed OpenAI's GPT-4o in key benchmark tests, including CCBench and OCRBench. The model also exhibits impressive multimodal capabilities, handling images, audio, and video. Furthermore, its text performance surpassed that of DeepSeek's V3 model in several areas, placing it on par with OpenAI's GPT-4.5.
Baidu, which introduced China’s first LLM in March 2023, is facing increasing pressure from domestic competitors such as DeepSeek, Alibaba, Tencent, and Bytedance, who have expanded rapidly in the AI space. The launch of Ernie 4.5 and Ernie X1 aims to help Baidu maintain its dominance in China’s highly competitive AI market.
While Baidu did not provide detailed benchmark data for Ernie X1, the company claims that the model performs similarly to DeepSeek’s R1 reasoning model but at a significantly lower cost. The pricing for accessing Ernie X1’s API is set at 2 yuan (US$0.28) per million token inputs and 8 yuan per million token outputs, offering a more affordable alternative to DeepSeek’s R1, which charges US$0.55 per million token inputs and US$2.19 per million token outputs.
Baidu’s CEO, Robin Li Yanhong, confirmed that Ernie 4.5 would be open-sourced by June 30, a strategic move to boost adoption of the model. Li noted that open-sourcing top-performing models, like DeepSeek’s, has proven to be an effective way to foster widespread use.
Despite these technological advancements, Baidu continues to face challenges in its core advertising business. The company reported a 2% year-on-year decline in revenue for the fourth quarter, and a 1% drop in total annual revenue, primarily due to a weak advertising market.