Inference
EN
LG AI Research

Exaone 4.0 32B

Exaone 4.0 32B is a reasoning model from LG AI Research in the Exaone 4.0 family. 8 benchmarks count toward its index, in 6 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2025-07
RankedEnough results for an index.
IndexAll categories, out of 100.32.4±8.7Median 51.0
CoverageIndex weight that has results.70%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.128KMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

32B parameters
16-bit
68.2 GB
8-bit
36.7 GB
4-bit
21.4 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 24 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.1 counted5.5—
CodingCode writing and repair.None counted——
ReasoningLogic problems and puzzles.1 counted36.8—
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.3 counted37.2215 / 268
MultilingualTasks in many languages.1 counted33.3—
InstructionTasks with strict prompt rules.1 counted20.3—
MathMath problems.1 counted45.3—

Results

8 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
EuroEval SpanishMultilingual48.3% ±0.752 / 176
EuroEval PortugueseMultilingual52.4% ±0.662 / 182
EuroEval ItalianMultilingual49.7% ±0.764 / 174
EuroEval GermanMultilingual41.8% ±0.673 / 191
EuroEval FrenchMultilingual53.9% ±0.670 / 178
Massive Multitask Language Understanding ProfessionalKnowledge81.8%34 / 75
American Invitational Mathematics Examination 2025Math85.3%26 / 53
EuroEval PolishMultilingual41.3% ±0.6106 / 209
EuroEval SwedishMultilingual41.0% ±1.0101 / 195
EuroEval DutchMultilingual41.2% ±0.697 / 175
Critical Physics TasksReasoning0.0%150 / 214
Artificial Analysis GPQA DiamondKnowledge62.8%152 / 199
Artificial Analysis Humanity's Last ExamKnowledge5.0%176 / 222
Artificial Analysis IFBenchInstruction33.5%127 / 146
τ²-Bench Tool-Agent-User EvaluationAgentic4.1%147 / 149

Sources

BenchLMCC BY-NC 4.0EuroEvalMIT

History

5 releases
Index31.9 → 32.4
23 Oct 202612 Nov 2026

Questions

How good is Exaone 4.0 32B?
Its index is 32.4 out of 100. That is place 245 of 283 ranked models.
What is Exaone 4.0 32B best at?
Knowledge: 37.2 out of 100, place 215 of 268.
What is the context window of Exaone 4.0 32B?
128K tokens.
Is Exaone 4.0 32B open source?
Yes. Its weights are published.
When was Exaone 4.0 32B released?
July 2025.

Badge

Exaone 4.0 32B on Inference360: place 245 of 283, index 32.4
HTML
<a href="https://inference360.ai/models/exaone-4-0-32b"><img src="https://inference360.ai/badges/exaone-4-0-32b.svg" alt="Exaone 4.0 32B on Inference360: place 245 of 283, index 32.4" height="20"></a>
Markdown
[![Exaone 4.0 32B on Inference360: place 245 of 283, index 32.4](https://inference360.ai/badges/exaone-4-0-32b.svg)](https://inference360.ai/models/exaone-4-0-32b)

Compare with

Nemotron 3 Nano 30B32.9Granite 4.2 3B33.2Magistral Small31.4K-Exaone42.3

More from LG AI Research

All models
Exaone 4.0 1.2BUnrankedK-Exaone42.3K-EXAONE 2.048.8