GLM-4-9B
| Property | Value |
|---|---|
| Author | THUDM |
| Parameters | 9 Billion |
| Context Length | 8K tokens |
| Model Type | Large Language Model |
| License | Custom (See LICENSE) |
| HuggingFace | THUDM/glm-4-9b-hf |
What is glm-4-9b-hf?
GLM-4-9B is the latest generation open-source pre-trained model in the GLM-4 series by Zhipu AI. It's a powerful 9B parameter model that demonstrates superior performance compared to Llama-3-8B across various benchmarks including semantics, mathematics, reasoning, and code generation tasks.
Implementation Details
The model is implemented using the Transformers library (requires version ≥4.46.0) and can be easily deployed using PyTorch. It supports both CPU and GPU inference, with optimal performance achieved through bfloat16 precision and automatic device mapping.
- Supports 26 languages including Japanese, Korean, and German
- 8K context length in base version
- Compatible with latest Transformers library
- Offers efficient inference with bfloat16 precision
Core Capabilities
- Strong performance on MMLU (74.7%)
- Exceptional C-Eval results (77.1%)
- Advanced mathematical reasoning (GSM8K: 84.0%)
- Superior code generation (HumanEval: 70.1%)
- Multi-language support across 26 languages
- Extended context handling capabilities
Frequently Asked Questions
Q: What makes this model unique?
GLM-4-9B stands out for its exceptional performance despite its relatively compact size. It outperforms larger models in specific tasks and offers comprehensive multilingual support, making it particularly valuable for diverse applications.
Q: What are the recommended use cases?
The model excels in various applications including mathematical reasoning, code generation, multilingual text processing, and general language understanding tasks. It's particularly suitable for applications requiring strong performance in academic and technical domains.
