LG AI Research has opened its most powerful AI model yet to outside developers — for free. The South Korean lab announced on Monday at Nvidia AI Day Seoul 2026 that the application programming interface (API) for K-Exaone 2.5, its frontier-class foundation model, is now available at no charge through LG's own platform.
Developers, startups, companies and research institutions can obtain an API key and use the model for service development and research, LG said. The announcement was made at the two-day Nvidia event at COEX in Seoul, where around 2,000 developers, researchers and corporate officials have gathered to share work on agentic AI, physical AI and large-scale AI infrastructure.
A 750-billion-parameter model, free to try
K-Exaone 2.5 is a technology validation model built by applying additional training to the second-generation proprietary model LG released in August. According to LG's model card, it has 750 billion parameters, with 31 billion active at any one time, and handles text in Korean and English with a 512,000-token context window — enough to take in entire codebases or book-length documents in a single pass.
The endpoint is OpenAI-compatible, so existing client code can point at it with the model name k-exaone_v2.5 and an API key from LG's platform. LG says the model now matches DeepSeek V4.1 Flash and GLM 5.3 on agentic benchmarks, although those claims have not yet been independently verified. LG has not said how long the free period will last or what limits will apply once it ends.
Sovereign AI, open to the world
The release is the latest step in South Korea's push to build homegrown frontier AI. LG's earlier K-Exaone model ranked seventh worldwide and first in Korea on a global open-weight leaderboard compiled by the evaluation group Artificial Analysis, and the company says it developed the model in-house over five years. The August release, K-EXAONE 2.0, was the country's largest open-source foundation model, with its full weights published under an Apache 2.0 licence.
LG is also pursuing closer cooperation with Nvidia. The lab said it is reviewing the adoption of Nvidia's Dynamo technology to improve inference processing efficiency and service scalability as usage of the free API grows.
The road to a January frontier model
Im Woo-hyung, head of LG AI Research, said the team is now focused on developing the third version of the model. "We hope that more developers and companies will experience the improved performance through this API release," he said. "We will concentrate on completing a frontier-level Exaone model by January next year."
The free API gives developers outside Korea an easy way to try a non-American, non-Chinese frontier contender — and it puts LG's work directly in front of the developers gathered at Nvidia's Seoul event, where Samsung, SK Hynix, LG Electronics and Naver Cloud are all presenting their own uses of Nvidia technology. For LG, the bet is that free access now turns into real-world validation before its full frontier model arrives in January.
