trurl-2-7B: benchmark results
VoiceLab's Trurl 2 7B, a Polish-and-English fine-tune of Llama 2 trained on 970,000 conversational samples (1.7 billion tokens) with a 4,096-token context, for commercial and research use. Provider: VoiceLab. Released 2023-08-16. Access: Open.
Aggregate Rating 1293 ± 11, rank #1494 of 1656 rated models, from 105 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - Econometrics | 33.33 | Accuracy (%) | 80.1 |
| MMLU-by-task - Anatomy | 50.37 | Accuracy (%) | 79.6 |
| MMLU-by-task - High School Physics | 34.44 | Accuracy (%) | 78.7 |
| MMLU-by-task - Formal Logic | 36.51 | Accuracy (%) | 76.5 |
| MMLU-by-task - Security Studies | 62.45 | Accuracy (%) | 74.9 |
| MMLU-by-task - High School Macroeconomics | 50.51 | Accuracy (%) | 70 |
| MMLU-by-task - Medical Genetics | 57 | Accuracy (%) | 68.4 |
| MMLU-by-task - High School Chemistry | 39.9 | Accuracy (%) | 67.9 |
| MMLU-by-task - High School Government And Politics | 76.68 | Accuracy (%) | 67.2 |
| MMLU-by-task - College Chemistry | 38 | Accuracy (%) | 66.4 |
| MMLU-by-task - Jurisprudence | 64.81 | Accuracy (%) | 66.2 |
| MMLU-by-task - Business Ethics | 53 | Accuracy (%) | 66.1 |
Interactive version: theaggregate.ai/model?slug=trurl-2-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.