LG AI Research

EXAONE enterprise models

Tier

Moderate

Headquarters

Seoul, South Korea

Founded

2020

Flagship

EXAONE

Models tracked

8

Artificial Analysis

Top intelligence

32.1

Artificial Analysis Index

Cheapest model

$0.00

blended /1M

Fastest output

0

tokens / sec

Flagship

LG AI Research’s top-scoring model on the Artificial Analysis Intelligence Index.

Most intelligent model

K-EXAONE (Reasoning)

Released 2025-12-31

Intelligence

32.1

Coding

27.0

Math

90.3

In $/1M

$0.00

Out $/1M

$0.00

Speed t/s

0

Benchmark breakdown

GPQA Diamond
78
Humanity’s Last Exam
13
SciCode
36
τ²-Bench (agentic)
74
Terminal-Bench Hard
23
IFBench
65

Models & benchmarks

Every LG AI Research model on file — intelligence, coding, price, and speed.

ModelReleasedIntelCodingMathIn $/1MOut $/1MSpeed t/sLatency s
K-EXAONE (Reasoning)2025-12-3132.127.090.3$0.00$0.0000.00
EXAONE 4.5 33B2026-04-0930.223.0$0.00$0.0000.00
K-EXAONE (Non-reasoning)2025-12-3123.413.544.0$0.00$0.0000.00
EXAONE 4.0 32B (Reasoning)2025-07-1516.714.080.0$0.00$0.0000.00
EXAONE 4.0 32B (Non-reasoning)2025-07-1511.79.439.3$0.00$0.0000.00
Exaone 4.0 1.2B (Reasoning)2025-07-158.33.150.3$0.00$0.0000.00
Exaone 4.0 1.2B (Non-reasoning)2025-07-158.12.524.0$0.00$0.0000.00
EXAONE 4.5 33B (Non-reasoning)2026-04-09$0.00$0.0000.00
Via Artificial Analysis

Trajectory

LG AI Research’s model releases over time, by intelligence.

Jul 2025 · 16.7▲ +13.5 indexApr 2026 · 30.2

House strengths

Where LG AI Research’s lineup over-indexes, averaged across its models.

Capability profile

Reasoning
47
Coding
28
Math
61
Agentic
23
Instruction
32

Best value

No pricing data to rank value.

Overview

The dedicated AI research arm of LG, producing the EXAONE family of bilingual enterprise and reasoning models.

Visit LG AI Research