Baidu unveiled a significant advancement in its Gen AI capabilities by launching a massive training cluster powered by 30,000 third-generation P800 Kunlun chips.
This infrastructure is designed to train models comparable to DeepSeek, handling hundreds of billions of parameters or enabling simultaneous fine-tuning for a thousand clients with billion-parameter models.
Baidu also introduced two new Gen AI models: Ernie 4.5 Turbo, a multimodal model excelling in coding and language comprehension, and Ernie X1 Turbo, a reasoning model. These developments underscore Baidu's commitment to enhancing AI applications across its ecosystem, including cloud services and content platforms.