gemma-7B (IT): benchmark results
Google's Gemma 7B instruction-tuned model, fine-tuned from the Gemma 7B pretrained checkpoint under the Gemma license, released February 2024; the linked card provides no further details here. Provider: Google. Released 2024-02-21. Access: Open.
Aggregate Rating 1292 ± 10, rank #1496 of 1656 rated models, from 299 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Trustworthy - Stereotype | 100 | Trust Score (%) | 93.8 |
| Fin-Bias | 97.6 | Average Herding Score (with rating) (self-reported) | 83.3 |
| Open CoT - LogiQA | 6.55 | CoT Gain (%) | 82.1 |
| Open PL LLM - EQ-Bench | 44.71 | EQ-Bench Score | 71.5 |
| Open LLM Leaderboard - MuSR | 42.74 | Score | 67.7 |
| LLM Trustworthy - Privacy | 83.69 | Trust Score (%) | 66.7 |
| LLM Trustworthy - Toxicity | 75.52 | Trust Score (%) | 66.7 |
| Enkrypt AI - Bias Risk | 79.59 | Risk Score | 63.5 |
| SALAD-Bench | 54.81 | Average Safety Score (%) | 60.6 |
| SALAD-Bench Base | 94.08 | Safety Score (%) | 60.6 |
| Open PL LLM - PSC (generative, 5-shot) | 91.98 | Binary F1 (%) | 60.1 |
| LLM Trustworthy - Fairness | 93.88 | Trust Score (%) | 58.3 |
Interactive version: theaggregate.ai/model?slug=gemma-7b-it · How It Works · Data refreshed daily, snapshot 2026-10-20.