产品功能
- - 提供各种大小的密集和混合专家(MoE)模型,包括0.6B、1.7B、4B、8B、14B、32B以及30B-A3B和235B-A22B。
- - 支持无缝切换思考和非思考模式,适用于不同的应用场景。
- - 显著提升了推理能力,在数学、代码生成和常识逻辑推理上超过了之前的QwQ(在思考模式下)和Qwen2.5指令模型(在非思考模式下)。
- - 具有卓越的人类偏好对齐能力,在创造性写作方面有突出表现。
Qwen3
Ali Qwen3: A Multilingual NLP Engine That Doubles Thinking and Creativity
Qwen3 is a new series of large language models developed by the Qwen team of Alibaba Cloud, with excellent natural language processing capabilities and wide applicability. It supports multiple languages and is suitable for scenarios such as text analysis, dialogue systems, and content generation. It provides dense and mixture-of-experts models of various scales, supports the switching between thinking and non-thinking modes, significantly improves reasoning ability, and performs outstandingly especially in mathematics, code generation, and commonsense logical reasoning. It has excellent human preference alignment ability and is suitable for c...