PLLuM is a family of large language models (LLMs) specialized in Polish and other Slavic/Baltic languages, with additional English data incorporated for broader generalization. Developed through an extensive collaboration with various data providers, PLLuM models are built on high-quality text corpora and refined through instruction tuning, preference learning, and advanced alignment techniques. These models are intended to generate contextually coherent text, offer assistance in various tasks (e.g., question answering, summarization), and serve as a foundation for specialized applications such as domain-specific intelligent assistants.
Key Highlights
Extensive Data Collection
We gathered large-scale, high-quality text data in Polish (around 150B tokens after cleaning and deduplication) and additional text in Slavic, Baltic, and English languages. Part of these tokens (28B) can be used in fully open-source models, including for commercial use (in compliance with relevant legal regulations).
Organic Instruction Dataset
We curated the largest Polish collection of manually created “organic instructions” (~40k prompt-response pairs, including ~3.5k multi-turn dialogs). This human-authored instruction set is based on an extensive typology of human-model interactions and it covers a range of subtle aspects of supervised fine-tuning (SFT) that might be overlooked with automated approaches (including large scale distillation of 'strong LLMs'). It was also designed to mitigate negative linguistic transfer from non-Polish textual data used in the pre-training phase.
Polish Preference Corpus
We created the first Polish-language preference corpus, featuring prompts and multiple model responses manually assessed by a demographically diverse team of annotators. This dataset teaches the model not only correctness (factual and linguistic) but also balance and safety—especially for potentially controversial or adversarial topics.
Evaluation Benchmarks
We developed custom benchmarks to evaluate our models on tasks relevant to Polish public administration, where PLLuM achieved top scores among all tested models. In broader Polish-language tasks, PLLuM models also attain state-of-the-art results.
Model Description
Note: this model was formerly denoted as PLLuM-12B-base, we have updated the name to make it conform to the naming convention of PLLuM models.
Below is a summary of the main PLLuM models, including their licenses, bases, and parameter sizes. All model names link to a specific Hugging Face resources, while the base models and licenses link to their respective sources or license references. Note that all -nc- models are intended to non-commercial use.
The models with fully open licenses are continuously pretrained on approximately 30 billion tokens of Polish text due to copyright considerations. The models with CC-BY-NC-4.0 licenses used approximately 150 billion tokens of Polish text. The models with the -nc and -chat suffix were aligned on human preferences and are generally safer and more efficient to use in dialog, general purpose scenarios.
Pretraining: All models were pretrained or continued-pretrained on large-scale Polish corpora (up to 150B tokens) plus a range of additional Slavic/Baltic and English texts.
Instruction Fine-Tuning: We refined the models on manually curated Polish “organic instructions” (approx. 40k), converted instructions from premium Polish corpora (approx. 50k), and synthetic instructions generated by strong LLMs (approx. 10k).
Alignment and Preference Learning: Manually annotated preference data taught the models to produce safer, balanced, and contextually appropriate responses, even in adversarial or sensitive cases.
Domain-Specific Adaptations: Specialized RAG-based (Retrieval Augmented Generation) models were developed for tasks like public administration, demonstrating strong performance in complex information retrieval and question answering.
Intended Use Cases
General Language Tasks: Text generation, summarization, question answering, etc.
Domain-Specific Assistants: Especially effective for Polish public administration and legal or bureaucratic topics where domain-aware retrieval is required.
Research & Development: Building blocks for downstream AI applications in academic or industrial settings, where a strong command of the Polish language is essential.
How to Use
Each PLLuM model can be loaded via the Hugging Face Transformers library (or compatible frameworks). For RAG-based scenarios, pair the model with a relevant vector store or document retrieval system.
Below are some recommended steps and code snippets:
1. Installation
Make sure you have the latest versions of transformers and torch (or another compatible deep learning framework) installed:
pip install transformers accelerate torch
2. Loading the Model
Use the following example to load one of the PLLuM models:
python
1from transformers import AutoTokenizer, AutoModelForCausalLM
23model_name ="CYFRAGOVPL/PLLuM-12B-chat-2412"# Replace with the PLLuM model name of your choice4tokenizer = AutoTokenizer.from_pretrained(model_name)5model = AutoModelForCausalLM.from_pretrained(model_name)
3. Using bfloat16 (BF16)
If your hardware (e.g., newer GPUs) supports bfloat16, you can reduce memory usage and potentially speed up inference:
python
1import torch
2from transformers import AutoTokenizer, AutoModelForCausalLM
34model_name ="CYFRAGOVPL/PLLuM-12B-chat-2412"5tokenizer = AutoTokenizer.from_pretrained(model_name)67# Load model in bfloat16 precision8model = AutoModelForCausalLM.from_pretrained(9 model_name,10 torch_dtype=torch.bfloat16,11 device_map="auto"# automatically places model layers on available devices12)
4. Generating an Example Text
python
12prompt ="Napisz krótki wiersz o wiośnie."# EN:"Write a short poem about spring."3inputs = tokenizer(prompt, return_tensors="pt").to(model.device)45outputs = model.generate(6**inputs,7 max_new_tokens=50,8 do_sample=True,9 top_k=50,10 top_p=0.9,11 temperature=0.712)1314generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)15print(generated_text)
5. Expected Output
Below is a sample (hypothetical) output for the prompt above:
css
1Przykładowy wiersz o tematyce wiosennej:23Wiosna, wiosna, wiosna, ach to ty!
4Kwiecień plecień wciąż przeplata,5trochę zimy, trochę lata.
6A ja nie mogę się już doczekać,7kiedy w kalendarzu ujrzę maj.
8Wtedy wszystko wkoło rozkwita,9a ptaki tak pięknie śpiewają.
10Wiosno, wiosno, czekam z utęsknieniem,11zrób mi tę przyjemność i przyjdź wreszcie, proszę!
Your results may vary depending on model parameters (e.g., temperature, top_k, top_p), hardware, and other settings.
6. Retrieval Augmented Generation (RAG)
Our Llama-PLLuM models (both chat and instruct versions) were additionally trained to perform well in Retrieval Augmented Generation (RAG) setting. The prompt is in .jinja format, where docs is a list of document texts and question is a query that should be answered based on the provided documents. If there is no answer in the provided documents model generates "Nie udało mi się odnaleźć odpowiedzi na pytanie".
Prompt:
Numerowana lista dokumentów jest poniżej:
---------------------
<results>{% for doc in docs %}
Dokument: {{ loop.index0 }}
{{ doc }}
{% endfor %}</results>
---------------------
Odpowiedz na pytanie użytkownika wykorzystując tylko informacje znajdujące się w dokumentach, a nie wcześniejszą wiedzę.
Udziel wysokiej jakości, poprawnej gramatycznie odpowiedzi w języku polskim. Odpowiedź powinna zawierać cytowania do dokumentów, z których pochodzą informacje. Zacytuj dokument za pomocą symbolu [nr_dokumentu] powołując się na fragment np. [0] dla fragmentu z dokumentu 0. Jeżeli w dokumentach nie ma informacji potrzebnych do odpowiedzi na pytanie, zamiast odpowiedzi zwróć tekst: "Nie udało mi się odnaleźć odpowiedzi na pytanie".
Pytanie: {{ question }}
Training Procedure
Datasets: ~150B tokens from Polish and multilingual sources, with ~28B tokens available for fully open-source commercial use.
Hyperparameters: Vary based on model size, typically including Adam or AdamW optimizers, a range of batch sizes, and carefully tuned learning rates.
Hardware & Duration: Training using Bem2 HPC (up to 300xH100 GPUs). Each model’s training time depends on parameter size and hardware configuration (~8 to ~25 days on multi-GPU cluster for 8B–70B sizes).
Evaluation and Benchmarks
Public Administration: PLLuM models demonstrated top-tier performance in specialized tasks relevant to government services.
Polish Language Tasks: Across a variety of internal benchmarks and standard corpora, PLLuM consistently outperforms other models in accuracy, coherence, and safety metrics.
Custom Tests: A unique preference corpus and alignment tests ensure robust, safe, and contextually accurate responses.
Limitations and Bias
Potential Hallucinations: Like other LLMs, PLLuM may occasionally produce factually incorrect or fabricated content.
Sensitivity & Bias: While extensive preference learning has been done, biases might still emerge, especially in controversial or subjective topics.
Context Length: Very long context tasks may challenge certain models, depending on memory constraints.
Ethical Considerations
PLLuM models are designed for constructive and responsible usage. Users should exercise caution when deploying them in production scenarios, especially for sensitive or regulated domains. Despite efforts to minimize harmful outputs, there is always a risk of generating offensive, biased, or inappropriate text. Human oversight and due diligence are advised.
Citation
If you use PLLuM models or any part of this repository in your research or deployment, please cite as follows (BibTeX):
@article{kocon2025pllum,
title = {PLLuM: A Family of Polish Large Language Models},
author = {
Jan Koco\'{n} and Maciej Piasecki and Arkadiusz Janz and Teddy Ferdinan and
\L{}ukasz Radli\'{n}ski and Bart\l{}omiej Koptyra and Marcin Oleksy and
Stanis\l{}aw Wo\'{z}niak and Pawe\l{} Walkowiak and Konrad Wojtasik and Julia
Moska and Tomasz Naskr\k{e}t and Bartosz Walkowiak and Mateusz Gniewkowski
and Kamil Szyc and Dawid Motyka and Dawid Banach and Jonatan Dalasi\'{n}ski
and Ewa Rudnicka and Bart\l{}omiej Alberski and Tomasz Walkowiak and
Aleksander Szcz\k{e}sny and Maciej Markiewicz and Tomasz Berna\'{s} and
Hubert Mazur and Kamil \.{Z}yta and Mateusz Tykierko and Grzegorz Chodak and
Tomasz Kajdanowicz and Przemys\l{}aw Kazienko and Agnieszka Karli\'{n}ska and
Karolina Seweryn and Anna Ko\l{}os and Maciej Chrab\k{a}szcz and Katarzyna
Lorenc and Aleksandra Krasnod\k{e}bska and Artur Wilczek and Katarzyna
Dziewulska and Paula Betscher and Zofia Cie\'{s}li\'{n}ska and Katarzyna
Kowol and Daria Miko\'{s} and Maciej Trzci\'{n}ski and Dawid Krutul and Marek
Koz\l{}owski and S\l{}awomir Dadas and Rafa\l{} Po\'{s}wiata and Micha\l{}
Pere\l{}kiewicz and Ma\l{}gorzata Gr\k{e}bowiec and Maciej Kazu\l{}a and
Marcin Bia\l{}as and Roman Roszko and Danuta Roszko and Jurgita
Vai\v{c}enonien\.{e} and Andrius Utka and Pawe\l{} Levchuk and Pawe\l{}
Kowalski and Irena Prawdzic-Jankowska and Maciej Ogrodniczuk and Monika Borys
and Anna Buli\'{n}ska and Wiktoria Gumienna and Witold Kiera\'{s} and Dorota
Komosi\'{n}ska and Katarzyna Krasnowska-Kiera\'{s} and \L{}ukasz
Kobyli\'{n}ski and Martyna Lewandowska and Marek \L{}azi\'{n}ski and
Miko\l{}aj \L{}\k{a}tkowski and Dawid Mastalerz and Beata Milewicz and
Agnieszka Anna Mykowiecka and Angelika Peljak-\L{}api\'{n}ska and Sandra
Penno and Zuzanna Przybysz and Micha\l{} Rudolf and Piotr Rybak and Karolina
Saputa and Aleksandra Tomaszewska and Aleksander Wawer and Marcin
Woli\'{n}ski and Joanna Wo\l{}oszyn and Alina Wr\'{o}blewska and Bartosz
\.{Z}uk and Filip \.{Z}arnecki and Konrad Kaczy\'{n}ski and Anna Cichosz and
Zuzanna Deckert and Monika Garnys and Izabela Grabarczyk and Wojciech
Janowski and Sylwia Karasi\'{n}ska and Aleksandra Kujawiak and Piotr Misztela
and Maria Szyma\'{n}ska and Karolina Walkusz and Igor Siek and Jakub
Kwiatkowski and Piotr P\k{e}zik
},
year = 2025,
journal = {arXiv preprint arXiv:2511.03823},
url = {https://arxiv.org/abs/2511.03823},
eprint = {2511.03823},
archiveprefix = {arXiv},
primaryclass = {cs.CL}
}
License
Different models within the PLLuM family are published under various licenses (Apache 2.0, CC-BY-NC-4.0, or Llama 3.1 license). Check each model’s entry in the table above for details.
Creators & Consortium
The PLLuM project is a unique collaboration between leading Polish scientific institutions and experts from various fields, working together to create a groundbreaking Polish language model. This research partnership combines diverse competencies and passions, forming a robust foundation for advancing AI in Poland.
We welcome feedback, collaboration, and further exploration of PLLuM models!
Acknowledgements
Project financed by the Minister of Digital Affairs under the targeted subsidy No. 1/WI/DBiI/2023: “Responsible development of the open large language model PLLuM (Polish Large Language Model) to support breakthrough technologies in the public and economic sector, including an open, Polish-language intelligent assistant for petitioners.”
