ABU DHABI, 3rd September, 2026 (WAM) -- The Institute of Foundation Models (IFM) at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) has launched K2 Horizon, a fleet of six AI foundation models ranging from 0.9 billion to 375 billion parameters.All six models are being released fully open, including weights, code, training data and methodology, enabling researchers and developers worldwide to inspect, reproduce and adapt them.According to MBZUAI, K2 Horizon delivers top-tier performance across reasoning, mathematics, coding and agentic tasks, with its 0.9B, 3.7B and 7B models achieving state-of-the-art performance at their respective scales.The 0.9B model is designed for constrained environments such as watches and glasses, while the 3.7B and 7B models target phones and other on-device applications. The 32B and 36B-A4B models support local hosting and on-premise deployment, while the flagship 375B-A23B is designed for demanding enterprise workloads.MBZUAI described K2 Horizon as the largest fully open model release in AI history, extending the K2 family's approach to transparency beyond the release of model weights.The six models largely share the same architecture, vocabulary, training methodology, interfaces and deployment tooling, allowing developers to move between models as their requirements scale.Eric Xing, Founder of IFM and President and University Professor of MBZUAI, said, “We believe meaningful AI progress depends on the ability to examine, build upon, and improve the technology, not simply access it through an API. The most important technology of our time should be built with the world, not kept from it.”K2 Horizon incorporates several technical innovations, including diffusion distillation, which IFM says can accelerate model performance by roughly three times without reducing response quality; a Mixture of Value Attention (MoVA) architecture designed to improve reasoning without additional computation; and dynamic model routing to direct tasks to the most cost-effective model.The 36B model activates only 4B parameters using MoVA, while the 375B flagship activates 23B parameters and is designed to compete with leading open-weight models in reasoning and agentic tasks.Hector Liu, Director of IFM’s Silicon Valley Research Lab, said, “Rather than release a single model, we’re releasing an entire fleet at once, six models, from one small enough to run on a watch to a flagship built for enterprise reasoning.”Established in May 2025, IFM operates from Abu Dhabi, Silicon Valley and Paris. MBZUAI is also training 80,000 UAE federal employees as agentic AI experts through a strategic knowledge partnership with the Federal Authority for Government Human Resources.K2 Horizon is available immediately through Hugging Face, vLLM and SGLang, while its API is available through inference partners including Compass, Cerebras, AWS and Nebius. The models and code are released under the Apache 2.0 licence.


(Kaynak: WAM.ae)