Kunlun Wanyi Releases TianGong Skywork-13B Series Large-scale Models, Commercial Use Barrier-free

2023-10-31

Kunlun Wanwei recently announced that its billion-level large language model "Tiangong" Skywork-13B series has been open-sourced. At the same time, it has also open-sourced the Chinese corpus dataset Skypile/Chinese-Web-Text-150B, which contains 600GB of data and 150 billion tokens, making it one of the largest open-source Chinese datasets available. In addition, the Skywork-13B series of large models will be fully available for commercial use without the need for developers to apply. The "Tiangong" Skywork-13B series currently includes two models with 13 billion parameters: the Skywork-13B-Base model and the Skywork-13B-Math model. They have surpassed other open-source large models such as LLaMA2-13B in several evaluation benchmarks, including CEVAL, CMMLU, MMLU, and GSM8K. In the field of Chinese text creation, they have performed well in areas such as technology, finance, government affairs, enterprise services, cultural and creative industries, and gaming. The open-source addresses for the Skywork-13B series of large models and the Chinese corpus dataset are as follows: Skywork-13B download address (Model Scope): https://modelscope.cn/organization/skywork Skywork-13B download address (Github): https://github.com/SkyworkAI/Skywork