Gemma 2 9B: benchmark results
Google's 9B Gemma 2 base model (June 2024), pretrained via knowledge distillation from the 27B teacher for above-class performance. Provider: Google. Released 2024-06-27. Access: Open.
Aggregate Rating 1457 ± 8, rank #958 of 1656 rated models, from 439 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LA Leaderboard - CoQCat | 78.51 | F1 (%) | 100 |
| LA Leaderboard - XNLI Basque | 53.46 | Accuracy (%) | 100 |
| LatentMD - Latent Failure Rate | 1.4 | Content-correct responses with a latent boundary failure (%) | 100 |
| Shlepa - Movies MC (RU) | 50.46 | Accuracy (%) | 100 |
| EuroEval Swedish NLU - Swerec | 80.82 | Matthews correlation coefficient (%) | 98.8 |
| LA Leaderboard - ARC Catalan | 67.87 | Accuracy (%) | 98.5 |
| LA Leaderboard - Belebele Basque | 79.44 | Accuracy (%) | 98.5 |
| LA Leaderboard - EpecKorrefBin Basque | 72.4 | Accuracy (%) | 98.5 |
| LA Leaderboard - EusProficiency Basque | 39.39 | Accuracy (%) | 98.5 |
| LA Leaderboard - EusReading Basque | 62.22 | Accuracy (%) | 98.5 |
| LA Leaderboard - MGSM Catalan | 20 | Exact Match (%) | 98.5 |
| LA Leaderboard - MGSM Galician | 22 | Exact Match (%) | 98.5 |
Interactive version: theaggregate.ai/model?slug=gemma-2-9b · How It Works · Data refreshed daily, snapshot 2026-10-20.