The Institute of Foundation Models (IFM) at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) has launched K2 Horizon, a suite of six fully open AI foundation models ranging from 0.9 billion to 375 billion parameters.
The UAE-developed models are released with their weights, source code, training data and development methodologies under the Apache 2.0 license. The release is designed to give developers a fully inspectable and reproducible alternative to closed AI systems.
The lineup covers deployment scenarios from wearable devices and smartphones to local servers and enterprise infrastructure. The smallest model is designed for edge devices such as smartwatches, while the 375B flagship targets complex reasoning and agentic workloads.
K2 Horizon uses diffusion distillation to generate token blocks in parallel, which the institute said can increase output speed by roughly three times. It also includes Mixture of Value Attention and dynamic model routing.
The models are available through Hugging Face, vLLM and SGLang, with API access offered by inference partners including Compass, Cerebras, AWS and Nebius. According to the announcement, the 0.9B, 3.7B and 7B models set new performance benchmarks in their size categories across mathematics, reasoning and coding.
Source: Middle East AI News
