GLM-130B: An Open Bilingual Pre-trained Model

Overview

GLM-130B is an open-source bilingual (English and Chinese) pre-trained language model with 130 billion parameters. It uses a General Language Model (GLM) architecture that combines autoregressive and autoencoding objectives. The model is designed for research and development, with its weights and code released to the community.

Best for

Best for
Researchers and developers needing an open, large-scale bilingual model for experimentation and benchmarking

Use cases

• Fine-tuning for bilingual text generation and understanding tasks
• Benchmarking large-scale language model performance in English and Chinese
• Research into scaling laws and model architectures for bilingual NLP

Notes

GLM-130B is an open-source bilingual (English and Chinese) pre-trained language model with 130 billion parameters. It uses a General Language Model (GLM) architecture that combines autoregressive and autoencoding objectives. The model is designed for research and development, with its weights and code released to the community.

Use cases

• Fine-tuning for bilingual text generation and understanding tasks
• Benchmarking large-scale language model performance in English and Chinese
• Research into scaling laws and model architectures for bilingual NLP

Pros

• Open-source weights and code enable full reproducibility and customization
• Bilingual capability supports both English and Chinese without separate models
• Large 130B parameter scale provides strong baseline for research

Cons

• Requires substantial computational resources for inference and fine-tuning
• Limited to research use; no commercial support or API
• Documentation and community resources are less mature than proprietary alternatives

Indexed from awesome-llm and enriched against its public facts.