RT by @huggingface: Building competitive AI from Korea for the global open-source community. SK…

TL;DR - SK Telecom released A.X K2, a 688B-parameter Mixture-of-Experts LLM with a 256K-token context window, published on Hugging Face under Apache 2.0. It signals a sovereign-AI push from Korea contributing a large, commercially usable open-weight model to the global ecosystem.

• Architecture: 688B total parameters using MoE sparse activation, positioning it among the largest openly released models.
• Context: 256K-token window, with the announcement highlighting math and long-context reasoning as areas of strength.
• Licensing: Apache 2.0, permitting commercial use and derivative work without restrictive terms.
• Caveat: the post is a promotional announcement — no benchmark numbers, active-parameter counts, or training details are provided, so the capability claims are unverified here.