GritLM-7B-KTO: benchmark results
GritLM's GritLM-7B-KTO, a KTO-aligned version of GritLM-7B, a generative representational model unifying text embedding and text generation in a single checkpoint. Provider: GritLM. Released 2024-04-16. Access: Open.
Aggregate Rating 1383 ± 17, rank #1279 of 1656 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Harmless | 65.37 | Normalized accuracy (%) | 79.8 |
| Open Korean LLM - Ko-IFEval | 36.95 | Strict accuracy, prompt/instruction mean (%) | 77 |
| Open Korean LLM - KorNAT-Helpful | 48.93 | Normalized accuracy (%) | 74.9 |
| Open Portuguese LLM - ASSIN2 RTE | 92.03 | Macro F1 (%) | 73.4 |
| Open Korean LLM - Ko-EQ-Bench | 22.43 | EQ-Bench score | 68 |
| Open Portuguese LLM - FaQuAD NLI | 72.71 | Macro F1 (%) | 66.7 |
| Open LLM Leaderboard - IFEval | 53.1 | Score | 63.1 |
| Open Korean LLM Leaderboard | 36.56 | Average Score (%) | 58 |
| Open Ko-LLM Leaderboard | 36.56 | Average (%) | 57.9 |
| Open Chinese LLM - TruthfulQA MC | 54.48 | Accuracy (%) | 57.1 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 22.21 | Exact match, flexible extract (%) | 54.6 |
| Open LLM Leaderboard - GPQA | 29.78 | Score | 53.6 |
Interactive version: theaggregate.ai/model?slug=gritlm-7b-kto · How It Works · Data refreshed daily, snapshot 2026-10-20.