ChatGLM大模型

智谱AI致力于打造新一代认知智能大模型，专注于做大模型的中国创新。公司于2020年底开始研发GLM预训练架构，2021年训练完成百亿参数模型GLM-10B，同年利用MoE架构成功训练出收敛的万亿稀疏模型，2022年合作研发了中英双语千亿级超大规模预训练模型GLM-130B。
2023年，智谱AI推出了基于千亿基座的对话模型ChatGLM，并开源单卡版模型ChatGLM-6B，使得研究者和个人开发者进行本地微调和部署成为可能。6月，ChatGLM2系列模型推出，提供丰富尺寸，适用于多种场景。与此同时，团队打造了AIGC模型及产品矩阵，包括AI提效助手智谱清言（chatglm.cn）、高效率代码模型CodeGeeX、多模态理解模型CogVLM和文生图模型CogView等。10月，在全新升级的ChatGLM3赋能下，生成式AI助手智谱清言已成为国内首个具备代码交互能力的大模型产品。
公司践行Mo

Zhipu AI is committed to building a new generation of cognitive intelligence large language models, focusing on Chinese innovations in large models. The company started developing the GLM pre-trained architecture at the end of 2020, completed the 10-billion-parameter model GLM-10B in 2021, and successfully trained a converged trillion-parameter sparse model using the Mixture of Experts (MoE) architecture in the same year. In 2022, it cooperated to develop the Chinese-English bilingual 130-billion-parameter super large pre-trained model GLM-130B.
In 2023, Zhipu AI launched the dialogue model ChatGLM based on the 100-billion-scale foundation model, and open-sourced the single-GPU version ChatGLM-6B, making it possible for researchers and individual developers to perform local fine-tuning and deployment. In June, the ChatGLM2 series of models was launched, providing diverse model sizes suitable for multiple scenarios. Meanwhile, the team built an AIGC model and product portfolio, including AI efficiency assistant Zhipu Qingyan (chatglm.cn), high-efficiency code model CodeGeeX, multimodal understanding model CogVLM, text-to-image model CogView, etc. In October, empowered by the newly upgraded ChatGLM3, the generative AI assistant Zhipu Qingyan has become the first domestic large model product with code interaction capabilities.
The company practices Mo