AI

SK Telecom Unveils Mobile AI Innovations with AX 3.1 Lite

LinkedIn Google+ Pinterest Tumblr

SK Telecom, a leading South Korean carrier, has strategically advanced in artificial intelligence with the release of AX 3.1 Lite. This new model, a compact large language model (LLM) designed for mobile applications, is now available through the open-source AI hub, Hugging Face.

The AX 3.1 Lite, featuring 7 billion parameters, is developed to enhance mobile deployments and language-focused applications such as call assistants. Despite its more manageable size, it holds significant potential as its performance parallels that of AX 4.0 Lite, which boasts 72 billion parameters. This lightweight model scored 96% on the Korean language KMMLU2 test and 102% on the CLIcK3 test, which evaluates cultural comprehension.

SKT’s goal is to enable more efficient AI systems on mobile devices, crucially impacting industries where localized language processing is vital. The open-source distribution via Hugging Face seeks to foster community collaboration and feedback, further refining this technology.

The release of AX 3.1 Lite represents a part of the broader AI strategy of SK Telecom. The Korean telecom giant continues to invest in the realm of AI, focusing on substantial projects such as the AI Infrastructure Superhighway. This initiative comprises AI Data Centers (AIDCs), GPU-as-a-Service, and Edge AI, aiming to support burgeoning AI applications.

Construction of hyperscale AI data centers is underway across South Korea. These centers, each exceeding 100 megawatts in capacity, serve as pivotal nodes for AI processing and storage needs, catering to both domestic and international AI tasks. Additionally, the GPU-as-a-Service platform democratizes access to high-performance computing, allowing developers to leverage powerful resources, aiding in AI model development without heavy initial hardware investments.

Edge AI, another key component, enhances low-latency processing by integrating AI closer to data sources like mobile devices and IoT sensors. This approach improves real-time data analysis and decision-making processes across various sectors.

Write A Comment