Gemma 2 2B (IT): benchmark results
Instruction-tuned Gemma 2 2B checkpoint. Provider: Google. Released 2024-07-31. Access: Open.
Aggregate Rating 1368 ± 7, rank #1341 of 1656 rated models, from 694 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SeaEval - Dialogue - SAMSum (Zero-Shot) | 31.12 | Average ROUGE (0-100) | 100 |
| SeaEval - Dialogue - DialogSum (Zero-Shot) | 25.97 | Average ROUGE (0-100) | 95.7 |
| Open Korean LLM - KorNAT-Helpful | 52.53 | Normalized accuracy (%) | 93.3 |
| Open Korean LLM - KorNAT-Harmless | 68.36 | Normalized accuracy (%) | 92 |
| LA Leaderboard - ParafrasesGL | 54.42 | Accuracy (%) | 89.7 |
| LA Leaderboard - ClinDiagnosES | 62.78 | Semantic Answer Similarity (x100) | 89.6 |
| Enkrypt AI - Insecure Code Risk | 6.22 | Risk Score | 88.7 |
| Enkrypt AI - Safety Risk | 19.16 | Risk Score | 88.4 |
| Open Korean LLM - Ko-IFEval | 43.74 | Strict accuracy, prompt/instruction mean (%) | 88.1 |
| LA Leaderboard - NoticIA | 25.7 | ROUGE-1 (%) | 86.2 |
| LA Leaderboard - EpecKorrefBin Basque | 57.07 | Accuracy (%) | 85.3 |
| LA Leaderboard - AQuAS | 68.53 | Semantic Answer Similarity (x100) | 83.8 |
Interactive version: theaggregate.ai/model?slug=gemma-2-2b-it · How It Works · Data refreshed daily, snapshot 2026-10-20.