Bielik-11B-v2: benchmark results
SpeakLeash's 11B Polish-language base model, initialized from Mistral-7B-v0.2 and further trained on 400 billion tokens of curated Polish text with ACK Cyfronet AGH. Provider: SpeakLeash. Released 2024-08-26. Access: Open.
Aggregate Rating 1372 ± 18, rank #1329 of 1656 rated models, from 194 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open PL LLM - PolQA Closed Book (generative, 5-shot) | 85.05 | Levenshtein Similarity (%) | 98.7 |
| Open PL LLM - PolQA Open Book (generative, 5-shot) | 91.46 | Levenshtein Similarity (%) | 95.6 |
| Open PL LLM - PPC (multiple choice, 5-shot) | 79.3 | Accuracy (%) | 88.5 |
| Open Arabic LLM - Aratrust Offensive | 92.75 | Accuracy (%) | 88.2 |
| Open PL LLM - PPC (generative, 5-shot) | 79 | Accuracy (%) | 86.5 |
| Open PL LLM - 8TAGS (generative, 5-shot) | 78.11 | Accuracy (%) | 84.9 |
| Open PL LLM - 8TAGS (multiple choice, 5-shot) | 78.84 | Accuracy (%) | 84.3 |
| Open Portuguese LLM - ASSIN2 RTE | 92.76 | Macro F1 (%) | 83.1 |
| Open PL LLM - CBD (generative, 5-shot) | 34.37 | Macro F1 (%) | 80.5 |
| Open PL LLM - CBD (multiple choice, 5-shot) | 33.29 | Macro F1 (%) | 77.4 |
| Open PL LLM - PolEmo2-OUT (multiple choice, 5-shot) | 74.9 | Accuracy (%) | 77.3 |
| Open PL LLM - RAG | 69.34 | Average RAG Score (%) | 77.3 |
Interactive version: theaggregate.ai/model?slug=bielik-11b-v2 · How It Works · Data refreshed daily, snapshot 2026-10-20.