reasoning
EXAONE 4.0 (32B)
LG AI Research 32B model that unifies reasoning and non-reasoning modes; weights are under a non-commercial research licence.
Levelmodel
Classreasoning
Parameters32B
Context131,072
Released15 Jul 2025
Opennessopen weights restricted
Runs onnot in the notes
Statusactive
Benchmark Results
vendor-reported
AIME 2025 2025math
Reasoning modevendor-reported
85.3
050100
Score in %. Higher is better.
EXAONE 4.0 model card
GPQA Diamondknowledge and reasoning
Reasoning modevendor-reported
75.4
050100
Score in %. Higher is better.
EXAONE 4.0 model card
LiveCodeBench v6coding
Reasoning modevendor-reported
66.7
050100
Score in %. Higher is better.
EXAONE 4.0 model card
MMLU-Proknowledge and reasoning
Reasoning modevendor-reported
81.8
050100
Score in %. Higher is better.
EXAONE 4.0 model card
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).
R&D Topics
Identity
- Developer
- LG AI Research
- Family
- EXAONE
- Released
- 15 Jul 2025 sourcesannouncement page
- Weights published
- –
- Record checked
- 11 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.
Architecture and Size
- Architecture
- dense transformer 64 layers, GQA (40 query, 8 KV heads), hybrid local sliding-window and global attention at 3:1, QK-Reorder-Norm; unified reasoning and non-reasoning modes
- Total parameters
- 32B sources
- Active parameters
- –
- Checkpoint parameters
- –
- Note
- 30.95B without embeddings
- Context length
- 131,072 tokens
- Input
- text
- Output
- text
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).
Openness and Access
- Openness
- open weights restricted
- Licence
- EXAONE AI Model License Agreement 1.2 - NC sources
- Access routes
- open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.
Deployment
- Runs on
- not in the notes
- Footprint
- not in the notes
- Runtimes
- not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.
Indonesia
- Indonesian support
- not stated
- Regional languages
- not in the notes
- Jakarta hosting
- unknown
- Languages
- English, Korean, Spanish
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.
Lineage
The notes record no lineage for this model.
Industry Use Cases
No industry use case links this model.
Links
DeveloperLG AI Researchcompany, South Korea
LG AI Researchcompany, South KoreaModelEXAONE 4.0 (32B)LG AI Research, reasoning
EXAONE 4.0 (32B)LG AI Research, reasoningR&D TopicsReasoning via Large-Scale RL
Use CasesChinese Medical LLMsHealthcareUse cases of models in the same R&D topic; none is recorded for this model itself.
Who Runs Itnone recordedDeployments of these use cases; none names this model.
Models 1
Organisations 1
Sources
- arxiv.org/abs/2507.11407 https://arxiv.org/abs/2507.11407
- huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B
- huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B/blob/main/LICENSE https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B/blob/main/LICENSE
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "EXAONE 4.0 (32B)", https://aurion.semai.id/models/lg-ai-research-exaone-4-0-32b/, records of October 2026.