K-EXAONE 2.0

K-EXAONE 2.0 ranks first on 3 of the 20 benchmarks it has scores on.

LG AI Research Released 31 Jul 2026 Family K-EXAONE The lab's model card

Benchmarks with a score

Ranked first, out of five or more models

In the top three

LG AI Research Jul 2026

Open weights Flagship

No Epoch Capabilities Index score yet

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Benchmark	Score	Rank	Source
OpenAI MRCR	94.4%	#1 of 10	Model card
OpenAI MRCR v2	94.4%	#1 of 9	Model card
IFEval	92.4%	#15 of 126	Model card
MMLU-Pro	83.5%	#45 of 151	Model card
Ko-LongBench (LG)	89.6%	#1 of 3	Model card
MMMLU	86.6%	#22 of 62	Model card
IFBench	72.6%	#30 of 70	Model card
KMMLU-Pro	69.1%	#3 of 7	Model card
CLIcK	84.2%	#2 of 4	Model card
Humanity's Last Exam	18.3%	#104 of 147	Model card
AA-LCR (Artificial Analysis Long Context Reasoning)	56.2%	#23 of 31	Model card
Global MMLU Lite 3.0	86.6%	#3 of 3	Model card

Benchmark	Score	Rank	Source
PolyMath	71.3%	#5 of 18	Model card
GPQA Diamond	82.2%	#98 of 284	Model card
HRM8K-KSM	91.1%	#1 of 2	Model card
AIME 2026	92.3%	#24 of 34	Model card
IMO-AnswerBench	78.6%	#28 of 37	Model card
HMMT 2026	78.4%	#17 of 18	Model card

Benchmark	Score		Rank	Source
SWE-bench Verified	68.2%		#87 of 150	Model card
SciCode	40.1%		#29 of 47	Model card

Benchmark	Score	Rank	Source
Claw-Eval	80%	#1 of 14	Model card
GAIA	23.3%	#45 of 57	Official leaderboard
τ³-bench Banking	14.2%	#17 of 20	Model card
Terminal-Bench 2.1	43.8%	#49 of 53	Model card

Benchmark	Score		Rank	Source
KGC-Safety (LG)	99.8%		#1 of 2	Model card
ROK-FORTRESS	89.5%		#1 of 1	Model card

The benchmarks the lab chose to headline, with the scores it gave.

Every result in the longer documents the lab published with it. 2 of them aren't in the launch post.