Bielik-11B-v2.1-Instruct: benchmark results
SpeakLeash's Polish 11B instruct fine-tune of the Bielik v2 base (depth-upscaled Mistral 7B v0.2), sibling variant to v2.2 (August 2024). Provider: SpeakLeash. Released 2024-08-26. Access: Open.
Aggregate Rating 1532 ± 16, rank #673 of 1656 rated models, from 82 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open PL LLM - PolQA Open Book (generative, 5-shot) | 92.39 | Levenshtein Similarity (%) | 99.5 |
| Open PL LLM - RAG | 76.79 | Average RAG Score (%) | 99.5 |
| Open PL LLM - PolQA Reranking (multiple choice, 5-shot) | 85.63 | Accuracy (%) | 98.7 |
| Open PL LLM - DYK (generative, 5-shot) | 71.26 | Binary F1 (%) | 97.2 |
| Open Portuguese LLM - FaQuAD NLI | 82.73 | Macro F1 (%) | 95.2 |
| MT-Bench PL - Writing | 9.5 | Judge Score (0-10) | 94.9 |
| MT-Bench PL - Roleplay | 9.45 | Judge Score (0-10) | 93.9 |
| Open PL LLM - PolEmo2-IN (multiple choice, 5-shot) | 87.12 | Accuracy (%) | 93.8 |
| Open PL LLM - DYK (multiple choice, 5-shot) | 70.05 | Binary F1 (%) | 92.3 |
| Open PL LLM - Multiple Choice | 62.88 | Average Multiple-Choice Score (%) | 92.3 |
| Open PL LLM - 8TAGS (multiple choice, 5-shot) | 79.37 | Accuracy (%) | 92.2 |
| Open PL LLM - PolQA Closed Book (generative, 5-shot) | 83.28 | Levenshtein Similarity (%) | 92.1 |
Interactive version: theaggregate.ai/model?slug=bielik-11b-v2-1-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.