On September 9, 2025 (local time), the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) in the UAE, in partnership with G42, officially launched a cost-effective AI inference model dubbed "K2 Think." Built upon Alibaba’s open-source Qwen 2.5 framework, this model delivers performance that outstrips similar models 20 times its size, all while operating with just 32 billion parameters. Moreover, it has successfully undergone operational testing on Cerebras hardware. Hector Liu, who heads the Foundation Models Institute at MBZUAI, disclosed that the team achieved significant performance enhancements by employing a mix of techniques, including Chain-of-Thought (CoT) supervised fine-tuning and test-time scaling. Specifically tailored for mathematics and scientific applications, the model is engineered to expedite research processes for intricate problems and democratize access to AI technology.