SK Telecom Releases A.X 3.1 Language Model on Hugging Face
SK Telecom has announced the release of its new proprietary large language model, A.X 3.1, on the open-source community platform Hugging Face. The A.X 3.1 model, developed entirely from scratch, features 34 billion parameters and is designed to significantly improve coding and mathematical reasoning capabilities.
The A.X 3.1 model is part of SK Telecom's dual-track strategy to enhance its AI capabilities. This strategy includes the in-house-developed A.X 3 series, which emphasizes self-sufficiency, and the A.X 4 series, optimized for the Korean business environment through large-scale continual pre-training with Korean data.
In addition to the A.X 3.1, SK Telecom has open-sourced a total of four AI models, including two A.X 3.1 models (standard and light) and two A.X 4.0 models (standard and light). These models are available for both academic research and commercial use, providing users with a range of options depending on their specific needs and environments.
The A.X 3.1 model has demonstrated strong performance in benchmarks such as the Massive Multitask Language Understanding in Korean (KMMLU) and Cultural and Linguistic Intelligence in Korea (CLIcK), achieving approximately 88% and 90% of the performance of the larger A.X 4.0 model, respectively. This release marks a significant step in SK Telecom's efforts to lead in the field of AI development and application.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like Daily AI Brief or AI Programming Weekly.
Also, consider following us on social media:
Daily AI Brief
Daily report covering major AI developments and industry news, with both top stories and complete market updates
Whitepaper
Tensordyne Napier: What If One Rack Could Do the Work of Nine?
Tensordyne
This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.
Read moreYou may also like
Microsoft Reveals 2 Trillion Parameter MAI-Base-2 Model
IFM and Inception42 Launch Jais 3 Arabic AI Models
SuperX Launches AI Token Platform and Content Creation Apps
LILT Launches AURORA Multilingual AI Leaderboard
ModelBest Releases MiniCPM-V 4.7 for Phones and Tablets
Daily AI Brief: the AI news that matters, in your inbox.