LG AI Research has opened free access to the API for its K-EXAONE 2.5 foundation model. Developers, startups, companies and research institutes can now use the model in their services and research at no cost.
The lab announced the move on November 9 at NVIDIA AI Day Seoul 2026, held at COEX in Seoul. Anyone who wants to try the model can obtain an API key through the EXAONE API Platform.
A test model built on the sovereign AI project
K-EXAONE 2.5 is a technology validation model built by applying additional training to K-EXAONE 2.0. That earlier model was the second LG built under the Korean government’s sovereign AI foundation model project, which aims to build homegrown AI systems and reduce the country’s reliance on foreign models.
LG says internal analysis across 24 benchmarks showed overall performance improving by more than 70 percent compared with the model released in July. The lab plans to apply the training techniques validated through K-EXAONE 2.5 to its third round model.
“Internal analysis showed that overall performance across 24 benchmarks improved by more than 70 percent compared with the model released in July.”
Kim Yi-reun, head of the EXAONE Lab at LG AI Research
How it measures up
On the company’s own tests using open datasets, K-EXAONE 2.5 performed on par with leading Chinese open weight models DeepSeek V4.1 Flash and GLM 5.3 in agentic AI, where a model plans tasks and uses tools on its own to complete real work.
It scored 74.6 on AutomationBench, which measures task automation with software as a service tools, against 74.9 for DeepSeek and 71.6 for GLM. It scored 77.3 on Claw-Eval, 78.0 on BrowseComp and 59.2 on SWE-Bench Pro, which test workplace problem solving, web browsing and software development.
Those figures come from LG’s own testing and have not been independently reproduced. LG has not said how long the free access will last or whether it will publish usage limits.
What comes next
LG AI Research is also reviewing NVIDIA’s Dynamo technology to improve the efficiency of running the model. The lab aims to complete a frontier level EXAONE model by January next year.
“We will focus on completing an EXAONE model with frontier-level performance by January next year.”
Lim Woo-hyung, co-head of LG AI Research
