Moreh

About

Moreh, a pioneering South Korean artificial intelligence company founded in 2020, is rapidly carving out a niche in the world of generative AI and large language models. The company is renowned for its advanced AI infrastructure software, which supports the lifecycle of hyperscale AI systems. Central to Moreh's offerings is the MoAI platform, a full-stack solution that enhances AI model scalability and management, leveraging parallelization for efficient scaling. This platform is compatible with various machine learning frameworks like PyTorch and TensorFlow and works seamlessly with diverse hardware setups, prominently featuring AMD GPUs. A standout achievement of Moreh is the creation of the largest Korean language model to date, boasting 221 billion parameters. This model was trained on 1,200 AMD MI250 GPUs using the MoAI platform, serving as a testament to their capability in handling extensive model training and demonstrating the competitive edge of AMD GPUs in AI applications. The impressive performance of their language model secured Moreh a top spot in Hugging Face's global performance assessments, underscoring their proficiency in developing high-performance AI models. Moreh's commitment to innovation is reflected in their R&D efforts, which include filing patents related to neural networks and machine learning technologies. Their dedication to advancing AI technology is further evidenced by their publications on fine-tuning large language models and GPU virtualization. With a robust team comprising experts from esteemed institutions like Seoul National University and substantial investment backing, including a $22 million Series B round from AMD and KT, Moreh is positioned as a pivotal player in the AI industry. The company's plans to release open-source models and ambitious growth projections highlight their ongoing contribution to and collaboration within the AI community.

Model Families

MoMo