
EXAONE 4.0 32B
Korea's first open-weight hybrid model fusing an LLM with a reasoning engine.
About
EXAONE 4.0 is LG AI Research's hybrid AI model, unveiled July 15, 2025, and is the company's second major milestone of the year after EXAONE Deep in March. It integrates a large language model for fluent natural language handling with a reasoning model that can form and verify hypotheses, a design shared by only a handful of models worldwide at the time.
LG reported that the 32B expert variant scored 92.3 on MMLU-Redux and 81.8 on MMLU-Pro, with 85.3 on AIME 2025 mathematics, and claimed it surpassed leading open-weight models from the US, China, and France in its own benchmark comparisons. The model also passed written exams for six Korean state-certified professional licenses, including medical and customs exams. A 1.2B on-device variant was released alongside for smartphones, automotive systems, and robots.
Weights are released on Hugging Face as an open-weight model for research, academic, and educational purposes under LG's EXAONE license, which carries non-commercial restrictions that should be verified before commercial use. Commercial API access is available through LG's partnership with FriendliAI.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 118; Trending score 82/100 from researcher assessment.