Llama-Krikri-8B-Instruct: benchmark results
Greek-language instruct model from Athena RC's ILSP, continually pretrained from Llama 3.1 8B on ~110B mostly Greek tokens then aligned. Provider: Meta. Released 2025-02-07. Access: Open.
Aggregate Rating 1498 ± 38, rank #2388 of 10916 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FinBen - FinText | 44.54 | Normalized Score | 79.2 |
| Greek MMLU - Greek-Specific | 74.73 | Accuracy (%) | 72.8 |
| Greek MMLU - Humanities | 70.29 | Accuracy (%) | 72.8 |
| Greek MMLU - Social Sciences | 70.57 | Accuracy (%) | 72.8 |
| Open LLM Leaderboard - IFEval | 60.79 | Score | 72.8 |
| Greek MMLU | 66.47 | Accuracy (%) | 70.4 |
| Greek MMLU - Other | 64.11 | Accuracy (%) | 69.1 |
| Greek MMLU - STEM | 62.62 | Accuracy (%) | 65.4 |
| Open LLM Leaderboard - GPQA | 30.29 | Score | 59.3 |
| Open LLM Leaderboard - MATH Level 5 | 11.78 | Score | 53 |
| Open LLM Leaderboard - BBH | 50.47 | Score | 50.2 |
| Open LLM Leaderboard - MuSR | 40.8 | Score | 48.6 |
Interactive version: theaggregate.ai/model?slug=llama-krikri-8b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.