EXAONE 4.0 32B

LG AI Research

Verified against Artificial Analysis · Jul 13, 2026

Capability

percentile index

Price /M

$0 / $0

in / out

Context

—

Avg score

47.7

11 benchmarks

AA Intelligence

10.6

index

Weights

Closed

Released

Jul 15, 2025

About

EXAONE 4.0 32B is a proprietary model in the EXAONE family from LG AI Research. Benchmarked on 11 evals, averaging 47.7.

Benchmark scores · 11

avg 47.7 — higher is better

MATHartificial-analysis · 2026-07

97.7

MMLU-Proartificial-analysis · 2026-07

81.8

AIMEartificial-analysis · 2026-07

LiveCodeBenchartificial-analysis · 2026-07

74.7

GPQA Diamondartificial-analysis · 2026-07

73.9

IFBenchartificial-analysis · 2026-07

36.3

SciCodeartificial-analysis · 2026-07

34.4

τ²-benchartificial-analysis · 2026-07

17.3

AA Long-Context Reasoningartificial-analysis · 2026-07

Humanity's Last Examartificial-analysis · 2026-07

10.5

Terminal-Bench Hardartificial-analysis · 2026-07

3.8

Details

PublisherLG AI Research

FamilyEXAONE

Parameters—

License—

WeightsClosed

ReleasedJul 15, 2025

History

full ledger →

Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers

Artificial Analysis$0/M