Gemma 2 9B (IT): benchmark results
Instruction-tuned Gemma 2 9B checkpoint. Provider: Google. Released 2024-06-27. Access: Open.
Aggregate Rating 1498 ± 5, rank #788 of 1656 rated models, from 1268 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AWARE-US (Global Weighting) - S3 Any Repair | 30 | Car match (%; share of the 40 persona-grounded car-search di | 100 |
| AWARE-US (Local Weighting) - S1 Unique Repair | 46.3 | Car match (%; share of the 41 persona-grounded car-search di | 100 |
| AWARE-US (Pairwise Ranking) - S1 Unique Repair | 24.4 | Car match (%; share of the 41 persona-grounded car-search di | 100 |
| Belebele European - Finnish | 88.89 | Accuracy (%, 0-shot) | 100 |
| KorSET (Direct Request) - Information and Safety Harms | 4.96 | Attack success rate (%) of KorSET's Information & Safety Har | 100 |
| KorSET (Direct Request) - Unfair Representation | 15.52 | Attack success rate (%) of KorSET's Unfair Representation pr | 100 |
| LA Leaderboard - Belebele Basque | 81.78 | Accuracy (%) | 100 |
| LA Leaderboard - ClinDiagnosES | 66.38 | Semantic Answer Similarity (x100) | 100 |
| LA Leaderboard - EpecKorrefBin Basque | 76.32 | Accuracy (%) | 100 |
| LA Leaderboard - QNLI Basque | 66.81 | Accuracy (%) | 100 |
| LA Leaderboard - Spanish Law Exams | 47.9 | Accuracy (%) | 100 |
| LA Leaderboard - WiC Basque | 54 | Accuracy (%) | 100 |
Interactive version: theaggregate.ai/model?slug=gemma-2-9b-it · How It Works · Data refreshed daily, snapshot 2026-10-20.