EXAONE-3.0-7.8B-Instruct: benchmark results
LG AI's EXAONE-3.0-7.8B-Instruct, an instruction-tuned 7.8B model in the EXAONE 3.0 series, released under LG AI's EXAONE license. Provider: LG AI. Released 2024-08-07. Access: Open.
Aggregate Rating 1416 ± 28, rank #1146 of 1635 rated models, from 44 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Horangi 4 - Korean Hate Speech | 71 | Accuracy (%) | 96.6 |
| Open LLM Leaderboard - IFEval | 71.93 | Score | 85.6 |
| Open LLM Leaderboard - MATH Level 5 | 30.44 | Score | 82.1 |
| Open Portuguese LLM - FaQuAD NLI | 74.24 | Macro F1 (%) | 70.1 |
| Open LLM Leaderboard - MMLU-Pro | 35.77 | Score | 54 |
| Open Portuguese LLM - ENEM | 56.75 | Accuracy (%) | 40.8 |
| Open Portuguese LLM - ASSIN2 RTE | 87.28 | Macro F1 (%) | 40.1 |
| Open Portuguese LLM - OAB Exams | 38 | Accuracy (%) | 36.9 |
| Open Portuguese LLM - BLUEX | 40.33 | Accuracy (%) | 32.6 |
| Open LLM Leaderboard - BBH | 41.74 | Score | 27 |
| Horangi 4 - Ko-HalluLens (Nonexistent Entities) | 59 | Refusal rate (%) | 24 |
| Horangi 4 - Ko-Moral | 66 | Accuracy (%) | 24 |
Interactive version: theaggregate.ai/model?slug=exaone-3-0-7-8b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-05.