Gemma-2-9B-It-SPPO-Iter1: benchmark results
UCLA-AGI's first-iteration SPPO self-play preference-optimization research checkpoint of Gemma 2 9B It. Provider: Google. Released 2024-06-29. Access: Open.
Aggregate Rating 1585 ± 41, rank #1186 of 10916 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-EQ-Bench | 61.39 | EQ-Bench score | 99.5 |
| Open Korean LLM - Ko-IFEval | 68.14 | Strict accuracy, prompt/instruction mean (%) | 99.2 |
| Open Korean LLM Leaderboard | 51.81 | Average Score (%) | 98.3 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 68.76 | Exact match, flexible extract (%) | 97.4 |
| Open Korean LLM - KorNAT-CKA | 34.6 | Normalized accuracy (%) | 94.1 |
| Open Korean LLM - KorNAT-Helpful | 51.25 | Normalized accuracy (%) | 89.6 |
| Open Korean LLM - KorNAT-Harmless | 67.21 | Normalized accuracy (%) | 88.5 |
| Open LLM Leaderboard - GPQA | 34.48 | Score | 86 |
| Open Korean LLM - Ko-Winogrande | 65.19 | Accuracy (%) | 83.3 |
| Open LLM Leaderboard - BBH | 59.72 | Score | 81.4 |
| Open LLM Leaderboard - MMLU-Pro | 39.07 | Score | 70.8 |
| Open LLM Leaderboard - MuSR | 40.99 | Score | 50.4 |
Interactive version: theaggregate.ai/model?slug=gemma-2-9b-it-sppo-iter1 · How It Works · Data refreshed daily, snapshot 2026-11-08.