
Creator: Remigiusz Kinas
Basal 2.0 Max
A decision model for classification, routing and scoring: a choice from a list, yes/no or a score on a scale. Available in the demo and through the Fabryka API.
MODELS / POLISH COMMUNITY
Discover models from Polish creators and find the source material for your project.
01 / MODEL FAMILIES

Creator: Remigiusz Kinas
A decision model for classification, routing and scoring: a choice from a list, yes/no or a score on a scale. Available in the demo and through the Fabryka API.

Creator: SpeakLeash · ACK Cyfronet AGH
A Polish language model for conversation, text work and tool calling, available in the chat and through the Fabryka API.

Creator: Fibogacci · Muszka Owocowa
A compact Polish assistant for short conversations and instructions.

Creator: adamo1139 / cpral
A Polish language-model family developed for text generation and research.

Creator: Sławomir Dadas
A model for retrieving documents by the meaning of a Polish query.

Creator: Sławomir Dadas
A model for ordering retrieved documents by their relevance to a question.

Creator: shefowl
A compact model for generating Polish speech from text.

Creator: Aleksander Ogrodzki / OrisTeam
A Polish base language model for text continuation and research.

Creator: Arkadiusz Słota
A small Polish language model for studying data composition and training.

Creator: Katarzyna Majzel-Pośpiech
A small model for experiments with Polish tokenization and language-model training.

Creator: Fibogacci
A compact classifier for moderation categories in Polish text.

Creator: Folx.it
A text-normalization model for preparing numbers, dates and abbreviations for speech.

Creator: RadLab
A model for recognizing personal-information entities in Polish text.

Creator: Piotr / piotr-ai
An independently developed model series for conversation and text generation.

Creator: Machinekind
A Polish text-to-speech project with a compact speech model.

Creator: Szymon Kocur
A model for experiments with nineteenth-century Polish language and text continuation.

Creator: Krzysztof Ociepa / Azurro Team
A family of Polish base models used to investigate training from scratch.

Creator: TeeZee
A conversational adaptation of Qra using Polish instruction data.

Creator: TeeZee
An experimental Polish adaptation of the Orpheus speech-generation model.

Creator: Dawid Kiełbasa
An English-language micro-model developed by a Polish creator for architecture experiments.
02 / PEOPLE AND PROJECTS
Meet the authors, researchers and teams. See what they build, and go to their models and sources. We keep pseudonyms exactly as the creators publish them.

CREATOR / PROJECT
Develops Basal: decision models, a custom inference engine and variants that run on a range of hardware.

CREATOR / PROJECT
Develops Bielik: Polish language models built by the SpeakLeash community and trained on ACK Cyfronet AGH supercomputers.

CREATOR / PROJECT
Publishes Muszka — a small Polish assistant — and Muszka Guard, a separate safety classifier.

CREATOR / PROJECT
Two separate accounts publishing the Poziomka family and its successive stages. Authorship of the base model and the fine-tuning is described in each version's repository.

CREATOR / PROJECT
Builds Polish retrieval models, embeddings and rerankers. Also publishes tools for comparing retrieval quality.

CREATOR / PROJECT
Publishes Polish adaptations and distillations of speech models, including the compact Pocket TTS. The public profile operates under a pseudonym.

CREATOR / PROJECT
Aleksander Ogrodzki publishes small Polish base models along with experiments in architecture and training.

CREATOR / PROJECT
Develops the GoLLeM Polish base models. Successive versions document experiments with data and continued training.

CREATOR / PROJECT
Researches the training of small Polish models and the impact of the tokenizer. Publishes GoLLeM 45M and the experiment code.

CREATOR / PROJECT
The Folx.it team publishes models and tools for Polish speech.

CREATOR / PROJECT
A lab developing Polish NLP: retrieval, answering questions from text, and recognizing personal data.

CREATOR / PROJECT
Piotr develops the Polanka series: Polish text models and separate multimodal variants.

CREATOR / PROJECT
Publishes Matcha-TTS-PL, lightweight Polish text-to-speech models with synthetic voices that run via ONNX.

CREATOR / PROJECT
Creates Wieszcz-XIX — models and a corpus of historical Polish for research and the digital humanities.

CREATOR / PROJECT
Krzysztof Ociepa and Azurro Team document training the Polish APT models from scratch and develop the ALLaMo framework.

CREATOR / PROJECT
Publishes fine-tunes of conversational models and speech-synthesis adaptations, including Qra Chat and Orpheus-TTS PL.

CREATOR / PROJECT
A Polish creator experimenting with very small language models.