Qwen2.5 0.5B Instruct

Qwen2.5-0.5B-Instruct is a causal language model for chat, instruction-tuned from the Qwen2.5 series. It is designed for general-purpose conversational AI, demonstrating strong capabilities in instruction following, generating long texts (up to 8,192 tokens), understanding structured data such as tables, and producing structured outputs like JSON. The model supports a context length of up to 32,768 tokens and is resilient to diverse system prompts, making it suitable for role-play and conditional chatbots. Multilingual support covers over 29 languages including Chinese, English, French, Spanish, Portuguese, German, Italian, Russian, Japanese, Korean, Vietnamese, Thai, and Arabic. Key improvements over the previous Qwen2 generation include enhanced knowledge depth, significantly better performance in coding and mathematics, and improved long-context handling. The model achieves these gains through specialized expert training in reasoning and structured generation tasks. Architectural specifics: 0.49 billion total parameters (0.36B non-embedding), 24 layers, grouped-query attention with 14 query heads and 2 key/value heads, RoPE, SwiGLU activation, RMSNorm, and tied word embeddings. Context length is 32,768 tokens; maximum generation length is 8,192 tokens. Licensed under Apache 2.0. The model is part of the Qwen2.5 lineup, which builds on the Qwen2 foundation and open-weight philosophy, offering a compact yet capable choice for developers requiring efficient, instruction-tuned chat performance via API.