zai-org/chatglm-6b

ChatGLM-6B is an open-source conversational AI model developed by Tsinghua University's KEG Lab that supports both Chinese and English. It solves the problem of making a capable large language model accessible to individuals and small teams who lack access to expensive high-end GPU hardware. At the time of its release, most comparable chat models required tens of gigabytes of GPU memory to run, making them impractical on consumer hardware. ChatGLM-6B addressed this through quantization, a technique that reduces the precision of the model's numerical weights to shrink its memory footprint. At its lowest quantization level (INT4), the model can run with as little as 6 gigabytes of GPU memory, which puts it within reach of many gaming and workstation graphics cards. The model has 6.2 billion parameters and was trained on roughly 1 trillion Chinese and English tokens. It uses a training approach similar to ChatGPT, combining supervised fine-tuning and reinforcement learning from human feedback to make responses feel natural and aligned with human preferences. Developers can load and query it using the Hugging Face Transformers library with just a few lines of Python. The repository also supports parameter-efficient fine-tuning through a technique called P-Tuning v2, which lets developers adapt the model to specific tasks using far less GPU memory than full fine-tuning would require. You would use ChatGLM-6B if you need a self-hosted bilingual Chinese-English chat model that can run locally without cloud costs. It is especially useful for researchers, developers building internal tools, and anyone who wants full control over a conversational AI without relying on an external API. The primary tech stack is Python with PyTorch and the Hugging Face Transformers library.