
In a bold move within the artificial intelligence landscape, the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) has introduced a low-cost reasoning model known as K2 Think, positioning itself as a rival to established players like OpenAI and DeepSeek. This announcement marks a significant entry into the competitive AI arena, particularly following DeepSeek's earlier launch of its advanced reasoning model, R1, which made waves for its claimed superior performance at a lower training cost. K2 Think, with its 32 billion parameters, is considerably smaller than its competitors, such as DeepSeek's R1, which boasts a staggering 671 billion parameters. This new model builds upon Alibaba's open-source Qwen 2.5 architecture and is supported by hardware from Cerebras, a prominent AI chip manufacturer. The researchers assert that K2 Think achieves performance levels comparable to the leading models from both OpenAI and DeepSeek, despite its reduced size. The collaboration with G42, a well-known UAE-based AI firm backed by Microsoft, played a pivotal role in the development of K2 Think. The team utilized innovative techniques, including long chain-of-thought (CoT) supervised fine-tuning and test-time scaling, to enhance the model's reasoning capabilities. Hector Liu, director of MBZUAI's Institute of Foundation Models, emphasized the model's system-oriented approach, stating that improvements are continuously implemented rather than merely releasing a static model. As the AI race intensifies, the U.S. and China remain the primary leaders, with American tech giants like OpenAI having set the stage early on with their foundation models. However, DeepSeek's R1 has underscored China's rising status in the AI field. The UAE's ambition to become a global AI leader reflects its desire to diversify its economy and increase geopolitical influence, even as it contends with regional competitors like Saudi Arabia. Despite the advancements represented by K2 Think, challenges remain. The UAE's AI sector still lags behind the extensive capabilities of U.S. and Chinese companies, which have had a considerable head start. The intent behind K2 Think is not to create a chatbot akin to ChatGPT, but to address specific applications in areas like mathematics and science. Richard Morton, managing director at MBZUAI's Institute of Foundation Models, highlighted the model's potential to significantly accelerate problem-solving processes that typically require extensive human resources and time. Ultimately, MBZUAI's K2 Think could help democratize access to advanced AI technologies, particularly in regions lacking the necessary capital and infrastructure. Morton remarked, "What we're discovering is that you can do a lot more with less," pointing to the model's promise for broader applications across various fields.
Morgan Stanley has identified two software stocks that have seen significant declines and suggests they may be worth pur...
CNBC | Feb 09, 2026, 20:55
The Federal Communications Commission (FCC) is reportedly probing ABC's daytime talk show, The View, a move that has dra...
Ars Technica | Feb 09, 2026, 20:35
Waymo has taken a significant step in its journey toward launching a driverless taxi service in Nashville by removing th...
TechCrunch | Feb 09, 2026, 23:15
OpenAI has announced the initiation of ad testing for users on its Free and Go subscription tiers in the U.S. This Go pl...
TechCrunch | Feb 09, 2026, 20:25
Musicboard, a popular music discovery and recommendation app, is facing significant challenges, leaving its users increa...
TechCrunch | Feb 09, 2026, 19:15