Zhipu AI¶

Zhipu AI is dedicated to creating a new generation of cognitive intelligent LLMs, focusing on innovation in LLMs in China, and linking hundreds of millions of users in the physical world through cognitive LLMs.

Based on a complete model ecosystem and full-process technical support, Zhipu AI emphasizes the research and development of ultra-large-scale training models, launching the dialogue model ChatGLM. At the same time, it practices the market concept of Model as a Service (MaaS) by launching an open platform for LLM MaaS.

On March 14, 2023, ChatGLM 1.0 started an invitation-only beta test. ChatGLM references the design concept of ChatGPT, injecting code pre-training into the hundred billion parameter base model GLM-130B, and achieving human intention alignment through techniques such as supervised fine-tuning. It boasts advantages such as bilingual support, high precision, fast inference, reproducibility, and cross-platform compatibility.

Also released during this period was ChatGLM-6B, which has 6.2 billion parameters and supports bilingual conversations in Chinese and English. Although its scale is not as large as the hundred billion model, it significantly reduces inference costs.

On June 27, 2023, the second generation of ChatGLM was officially launched. Building on the smooth dialogue and low deployment threshold of the first-generation model, it introduces more powerful performance, allows for more rounds of dialogue and longer context, enables more efficient inference, and permits more open protocols.

On July 15, 2023, Zhipu AI announced that ChatGLM allows free commercial use.