Qwen/Qwen2-0.5B-Instruct is a 0.5 billion parameter instruction-tuned causal language model from the Qwen2 series, developed by Qwen. Built on a Transformer architecture with SwiGLU activation and an improved tokenizer, it demonstrates enhanced performance across language understanding, generation, multilingual capabilities, coding, mathematics, and reasoning. This model is designed for general-purpose instruction following, offering competitive capabilities in a compact size.
Loading preview...
Qwen2-0.5B-Instruct Overview
Qwen2-0.5B-Instruct is a 0.5 billion parameter instruction-tuned model from the Qwen2 series, developed by Qwen. This model is part of a new generation of Qwen LLMs, built on a Transformer architecture incorporating features like SwiGLU activation, attention QKV bias, and group query attention. It also utilizes an improved tokenizer designed for multiple natural languages and code.
Key Capabilities & Performance
Qwen2-0.5B-Instruct shows significant improvements over its predecessor, Qwen1.5-0.5B-Chat, across various benchmarks. It was pretrained on a large dataset and further refined with supervised finetuning and direct preference optimization. Notable benchmark improvements include:
- MMLU: 37.9 (vs. 35.0 for Qwen1.5-0.5B-Chat)
- HumanEval: 17.1 (vs. 9.1 for Qwen1.5-0.5B-Chat)
- GSM8K: 40.1 (vs. 11.3 for Qwen1.5-0.5B-Chat)
- C-Eval: 45.2 (vs. 37.2 for Qwen1.5-0.5B-Chat)
Use Cases
This model is suitable for a wide range of instruction-following tasks, leveraging its enhanced capabilities in:
- Language understanding and generation
- Multilingual applications
- Coding assistance
- Mathematical problem-solving
- General reasoning tasks
Its compact size (0.5B parameters) combined with a 32K context length makes it efficient for deployment in scenarios requiring a capable yet lightweight model.