3 Sources
[1]
China's Baidu says its Kunlun chip cluster can train DeepSeek-like models
BEIJING/SHANGHAI, April 25 (Reuters) - China's Baidu (9888.HK), opens new tab has successfully "illuminated" a cluster comprising 30,000 of its self-developed, third generation P800 Kunlun chips, which can support the training of DeepSeek-like models, its CEO said on Friday. Robin Li made the
[2]
Baidu founder highlights 'shrinking' demand for DeepSeek's text-based AI
Baidu's founder has said demand for the type of text-based models developed by generative AI sensation DeepSeek is "shrinking", as his search group seeks to reestablish itself as an artificial intelligence leader in China. In a striking criticism of the limitations of China's AI darling, Robin Li
[3]
China's Baidu says its Kunlun chip cluster can train DeepSeek-like models
China's Baidu has successfully "illuminated" a cluster comprising 30,000 of its self-developed, third generation P800 Kunlun chips, which can support the training of DeepSeek-like models, its CEO said on Friday. Robin Li made the announcement at the company's annual developer conference where the
Share
Copy Link
Baidu announces a powerful Kunlun chip cluster capable of training large AI models, introduces new multimodal AI models, and criticizes text-only AI limitations, signaling a shift in China's AI landscape.

Baidu, the Chinese search engine giant, has made a significant advancement in AI hardware capabilities. CEO Robin Li announced at the company's annual developer conference that Baidu has successfully "illuminated" a cluster of 30,000 third-generation P800 Kunlun chips
1
. This powerful cluster is capable of training AI models comparable to DeepSeek, with hundreds of billions of parameters, or simultaneously fine-tuning thousands of models with billions of parameters1
.Baidu unveiled two new multimodal AI models at the conference:
Ernie 4.5 Turbo: This model reportedly matches industry benchmarks in various tasks, including coding and linguistic comprehension
1
.Ernie X1 Turbo: A new reasoning model that expands Baidu's AI capabilities
1
.These models are not limited to text generation but also include audio, image, and video capabilities, positioning Baidu competitively in the multimodal AI space
2
.In a notable shift, Robin Li criticized the limitations of text-based AI models like those developed by DeepSeek. He stated that "the market for text models is shrinking," highlighting the need for more versatile, multimodal AI solutions
2
. Li also pointed out that DeepSeek's models are more prone to hallucinations and are slower and more expensive than other domestic offerings2
.Baidu emphasized the importance of AI applications, with Li stating, "There are many (AI) models, but it's apps that rule the world. The application is the king"
3
. The company plans to integrate its AI capabilities across various applications, including its cloud drive and content platform Baidu Wenku1
.Related Stories
These announcements come amid intense competition in China's AI market. Baidu's shares in Hong Kong rose more than 4% following the news
2
. The company faces stiff competition from other tech giants like Alibaba and emerging players like ByteDance's Doubao2
.Despite these advancements, Baidu's Ernie Bot has struggled to gain traction in the competitive Chinese AI market
1
. The company has recently shifted its strategy, making its models open-source and focusing on enterprise applications2
. With these new developments, Baidu aims to reestablish itself as a leader in China's rapidly evolving AI landscape.Summarized by
Navi
13 Nov 2025•Technology
10 Feb 2025•Technology

16 Mar 2025•Technology

1
Science and Research

2
Policy and Regulation

3
Technology