The Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) in Abu Dhabi has introduced a new low-cost reasoning model named K2 Think, aiming to compete with AI models from DeepSeek and OpenAI.
This announcement follows DeepSeek’s January claim of nearing OpenAI’s capabilities with significantly less budget and energy. K2 Think distinguishes itself by utilizing fewer parameters than its competitors while maintaining comparable performance in reasoning tasks.
MBZUAI describes K2 Think as “a new class of reasoning model” that leverages long chain-of-thought supervised fine-tuning and reinforcement learning with verifiable rewards to enhance logical depth and accuracy. Hector Liu, director of MBZUAI’s institute of foundation models, emphasized that K2 Think is approached as a dynamic system, continuously improved through deployment and iterative adjustments.
According to MBZUAI, K2 Think is among the fastest and most efficient reasoning systems available, processing approximately 2,000 tokens per second, equivalent to around 1,500 words. The model is built on Alibaba’s Qwen 2.5 large language model and operates on hardware provided by AI chipmaker Cerebras.
Mirroring DeepSeek’s R1 model, K2 Think is open source, making its training data and weights publicly accessible. This transparency, according to the university, will allow the global research community to scrutinize, replicate, and expand upon the model’s reasoning capabilities.
MBZUAI highlights that K2 Think represents a significant milestone for AI development in the UAE. The university believes it exemplifies how open innovation and public-private partnerships can establish Abu Dhabi as a prominent global AI leader, showcasing that ingenuity and collaboration are crucial factors in shaping the future of AI reasoning.
The introduction of K2 Think could influence the global AI landscape, demonstrating the UAE’s potential to compete with leading nations like the U.S. and China through strategic innovation and partnerships.
