EXAONE-4.0-32B (Thinking): benchmark results
Provider: LG AI. Released 2025-07-15. Access: Open.
Unified ELO 1566 ± 1, rank #867 of 3078 rated models, from 43 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Horangi 4 - KMMLU | 86 | Accuracy (%) | 77.9 |
| Horangi 4 - Ko-Moral | 80 | Accuracy (%) | 74 |
| Horangi 4 - Ko-HalluLens (Nonexistent Entities) | 82 | Refusal rate (%) | 69.2 |
| Horangi 4 - HAE-RAE Bench (Reading Comprehension) | 89 | Accuracy (%) | 64.9 |
| Horangi 4 - GLP - General Knowledge | 84.41 | Score (%) | 53.8 |
| Horangi 4 - BFCL | 67.05 | Accuracy (%) | 52.5 |
| Horangi 4 - GLP - Semantic Analysis | 76 | Score (%) | 51.4 |
| NOLLI - Cipher - English | 35.7 | Accuracy (%) | 50 |
| Horangi 4 - KMMLU-Pro | 68 | Accuracy (%) | 49.5 |
| Horangi 4 - Ko-TruthfulQA | 80 | Accuracy (%) | 49.5 |
| Horangi 4 - BigCodeBench (100-task subset) | 49 | Pass rate (%) | 48.6 |
| Horangi 4 - KoBALT-700 (Semantics) | 63 | Accuracy (%) | 48.1 |
Interactive version: theaggregate.ai/model?slug=exaone-4-0-32b-thinking · How It Works · Data refreshed daily, snapshot 2026-09-19.