Alibaba-NLP/WebDancer-32B

Alibaba-NLP/WebDancer-32B is a 32 billion parameter agentic search reasoning model developed by Alibaba-NLP, designed for autonomous information seeking. It utilizes a ReAct framework and a four-stage training paradigm to acquire autonomous search and reasoning skills. This model excels at complex web-based tasks, achieving strong performance on benchmarks like GAIA and WebWalkerQA.