SK Telecom and Krafton release A.X-K2, a 688B MoE sovereign model
SK Telecom and Krafton released A.X-K2, a 688-billion-parameter Mixture-of-Experts model with 33 billion active parameters, as part of South Korea's sovereign AI effort. The package also includes specialized variants aimed at local and enterprise open-weight use.
Key takeaway
A major Asian carrier-game studio partnership is putting a frontier-class MoE into the open-weight sovereign lane, expanding alternatives beyond Western lab releases.
Context
A.X-K2 is framed as a sovereign AI deliverable rather than an incremental Western lab update, pairing SK Telecom and Krafton behind a single large Mixture-of-Experts stack. At 688 billion total parameters with 33 billion active, it lands in the large open-weight class that local deployers and enterprises track for self-hosted capacity.
The release matters now because sovereign AI programs are competing on usable open weights, not only closed APIs. Specialized variants in the package suggest an intent to cover narrower workloads beyond a single general model, which is the practical bar for national and enterprise adoption.
Numbers to know
- 688BTotal parameters in A.X-K2
- 33BActive parameters in the MoE design