Huawei Announces Launch of "Pangu Large Model" AI in April 255

News from March 27: Huawei Cloud's official website shows that Huawei's Pangu series ai large model is about to officially launch. Currently, the NLP large model, CV large model, and scientific computing large model (meteorological large model) within the AI large model have all been marked as coming soon.

The Pangu large model consists of multiple large models, including the NLP large model, CV large model, multimodal large model, and scientific computing large model. Through model generalization, it solves the challenges of AI scaling and industrialization that traditional AI workshop-style development models cannot address. It can support a variety of natural language processing tasks, including text generation, text classification, Q&A systems, and more.

Specifically, the Pangu NLP large model uses the Encoder-Decoder architecture for the first time, balancing the understanding and generation capabilities of the NLP large model, ensuring embedding flexibility across different systems. In downstream applications, only a small number of samples and learnable parameters are needed to quickly fine-tune and adapt a hundred-billion-parameter large model to downstream tasks. This model has demonstrated strong performance in intelligent public opinion analysis and intelligent marketing.

1. Huawei's Pangu NLP, comparable to ChatGPT, also reaches the hundred-billion-parameter scale. It is the world's largest Chinese language AI training large model, jointly developed by Huawei and the Pengcheng Laboratory. Pengcheng Cloudbrain II provides the computing power foundation.

2. Computing power is the key to AI. Pengcheng Cloudbrain II ranked first globally in AI computing power in 2022 and can process hundred-billion-parameter models, providing the AI computing power foundation for domestic giants such as Huawei Pangu and Baidu NLP. Pengcheng Cloudbrain II is jointly built by SDG Information and Huawei.

On April 25, 2021, Huawei launched the Pangu NLP AI training large model, jointly developed by Huawei Cloud and Pengcheng Laboratory, with computing power support provided by Pengcheng Cloudbrain II. It is the industry's first hundred-billion-parameter Chinese NLP large model for generation and understanding.

According to the introduction on Huawei's official website, the Pangu NLP large model ranks first in both the overall leaderboard and the single category of reading comprehension, breaking three world records. It scored 83.046 on the overall leaderboard, with industry-leading scores across multiple subtasks, making it the pre-trained model currently closest to the human understanding level (85.61).

Huawei's ChatGPT equivalent was forged over three years. The Pengcheng Pangu large model is the industry's first hundred-billion-parameter Chinese NLP large model for generation and understanding, and is currently the model closest to the human understanding level.