EXAONE 4.5 33B

EXAONE 4.5 33B places in the top three on We-Math.

LG AI Research Released April 9th, 2026 Family EXAONE The lab's model card Compare with others

Benchmarks with a score

Ranked first, out of five or more models

In the top three

LG AI Research Apr 2026

EXAONE 4.5 33B

Open weights Mid-size

139.5 #180 of 404

Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.

MMMU-Pro 68.6% #55 /90
IFBench 62.6% #55 /80
AA-LCR (Artificial Analysis Long Context Reasoning) 50.6% #32 /38

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.

Knowledge and language

Benchmark	Score	Rank	Source
IFEval	89.6%	#38 of 135	Lab model card
MMLU-Pro	83.3%	#50 of 167	Lab model card
MMMLU	85.4%	#32 of 71	Lab model card
HallusionBench	63.7%	#18 of 31	Lab model card
KoBALT	52.1%	#2 of 3	Lab model card
IFBench	62.6%	#55 of 80	Lab model card
KMMLU-Pro	67.6%	#5 of 7	Lab model card
AA-LCR (Artificial Analysis Long Context Reasoning)	50.6%	#32 of 38	Lab model card

Math and science

Benchmark	Score	Rank	Source
AIME 2025	92.9%	#34 of 143	Lab model card
GPQA Diamond	80.5%	#116 of 302	Lab model card
AIME 2026	92.6%	#22 of 38	Lab model card

Coding

Benchmark	Score		Rank	Source
LiveCodeBench v6	81.4%		#28 of 90	Lab model card

Agents and tools

Benchmark	Score		Rank	Source
tau2-bench	72%		#56 of 79	Lab model card

Professional work

Benchmark	Score		Rank	Source
MedXpertQA-MM	42.1%		#16 of 21	Lab model card

Vision, audio and robotics

Benchmark	Score	Rank	Source
We-Math	79.1%	#3 of 13	Lab model card
MMMU	78.7%	#24 of 103	Lab model card
MathVista	85%	#18 of 70	Lab model card
BLINK	68.8%	#8 of 29	Lab model card
LogicVista	73.8%	#4 of 14	Lab model card
OCRBench v2	63.2%	#8 of 22	Lab model card
AI2D	89%	#23 of 52	Lab model card
OmniDocBench v1.5	81.2	#4 of 9	Lab model card
MATH-Vision (MathVision)	75.2%	#25 of 51	Lab model card
MMMU-Pro	68.6%	#55 of 90	Lab model card
MMStar	74.9%	#22 of 36	Lab model card
CharXiv	71.7%	#54 of 75	Lab model card
ChartQA Pro	62.2%	#7 of 8	Lab model card
BabyVision	18.8%	#25 of 26	Lab model card
K-Viscuit	80.1%	#1 of 1	Lab model card
KMMMU	42.7%	#1 of 1	Lab model card
KRETA	91.9%	#1 of 1	Lab model card

Its lab's numbers put it at 139.5, number 180 of 404

The Self-Reported Index fits one score to the 25 capability results LG AI Research reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 137.8 to 141.2. How we compute it.