LG Unveils Korea's Largest AI Model, K-EXAONE 2.0, Open-Sourced to Rival Chinese Tech
LG AI Research Institute has officially released K-EXAONE 2.0, the largest AI foundation model ever built in South Korea. The model, available on Hugging Face since July 31, is the second generation under the Sovereign AI Foundation Model Project backed by Korea's Ministry of Science and ICT. With a hybrid attention MoE architecture, it packs a whopping 750 billion total parameters—37 billion activated—making it more than three times the size of its predecessor. This isn't just a step up; it's a giant leap.
What's turning heads is that K-EXAONE 2.0 is fully open-sourced under the permissive Apache 2.0 license. That's a rarity in Korea's AI scene, where proprietary models have been the norm. By opening the doors, LG is signaling a bold shift toward collaboration and transparency.

Performance that speaks for itself
Numbers don't lie, and K-EXAONE 2.0's are impressive. Across 24 benchmark tests, it averaged 70.1 points—a jump of over 10% from the first generation's 63.3. In coding and agent coding tasks, the improvement hit around 30%. That's not just incremental; it's transformative.
But the real story is how it stacks up against the competition. In long-context understanding, K-EXAONE 2.0 outshines Zhipu's GLM-5.1. When it comes to agent tool usage, it surpasses both GLM-5.1 and Qwen3.5. And in instruction execution, it's neck-and-neck with the likes of GLM-5.1, DeepSeek V4 Pro Max, and Qwen3.5. These are the heavyweights of the open-source world, and Korea's new kid on the block is holding its own.
A new player in the East Asian AI race
This launch isn't happening in a vacuum. The global open-source AI arena is heating up, with China's Kimi K3 sparking a wave of releases. Now, South Korea is throwing its hat in the ring under the banner of "sovereign AI." The message is clear: East Asia is no longer just a spectator in the AI revolution—it's a key player.
The triangular competition among China, Japan, and South Korea is taking shape, and it's anyone's game. With K-EXAONE 2.0, LG isn't just catching up; it's setting the pace. The question now is: what will the next move be?
Key Points
- K-EXAONE 2.0 is South Korea's largest AI model, with 750 billion parameters (37B active).
- Fully open-sourced under Apache 2.0, a first for Korea's AI industry.
- Outperforms Chinese models like GLM-5.1 and Qwen3.5 in key benchmarks.
- Signals South Korea's entry into the global open-source AI competition.