Qwen2.5 Math
About
The Qwen2.5-Math family consists of large language models (LLMs) adept at solving mathematical problems in English and Chinese 21012. These models employ advanced techniques like Chain-of-Thought (CoT) and Tool-Integrated Reasoning (TIR) to enhance their analytical abilities 21012. The lineup includes both base models and instruction-tuned versions across various scales (1.5B/7B/72B), as well as a 72B reward model that optimizes training 2812. Contrast to its predecessor, Qwen2-Math, this series offers bilingual support and demonstrates superior performance on mathematical benchmarks 2910. Available on platforms like Hugging Face and ModelScope, these models are dedicated to mathematical tasks and are not suitable for other applications 2712.