Qwen3 is the latest generation of large language models in the Qwen series, offering both dense and mixture-of-experts models. Qwen3-32B is a 32.8 billion parameter causal language model with a native context length of 32,768 tokens that can be extended to 131,072 tokens using YaRN scaling techniques. The model features a unique capability to seamlessly switch between thinking mode for complex logical reasoning, mathematics, and coding tasks, and non-thinking mode for efficient general-purpose dialogue. This switching can be controlled through the enable_thinking parameter or through user commands like /think and /no_think in conversations. Qwen3-32B significantly enhances reasoning capabilities, surpassing previous QwQ and Qwen2.5 instruct models on mathematics, code generation, and commonsense reasoning, while also excelling in human preference alignment for creative writing, role-playing, and multi-turn dialogues. The model supports over 100 languages and dialects with strong multilingual instruction-following and translation capabilities. It also demonstrates strong expertise in agent capabilities, enabling precise tool integration in both thinking and non-thinking modes, making it well-suited for complex agent-based tasks.