Bielik is one of Poland\u2019s AI language models which, unlike foreign alternatives, is specialised in the Polish language. Its key advantage for domestic users is that it can be run locally, on their own infrastructure, without the need to send data outside the organisation. The creators of Bielik have announced the release of a new version in June, which will support 50 languages, including Chinese and Arabic.<\/p>\n
\u201cInterest in the Bielik project is quite high. Why? First of all, it is a small, compact model, and that is its advantage. We can run it on our own infrastructure, and wherever state, banking or medical secrecy is involved, there is a need to use models on proprietary infrastructure. Bielik is a response to this need,\u201d Sebastian Kondracki, CEO of Bielik.AI, tells Newseria.<\/p>\n
As the Ministry of Digital Affairs points out, many AI tools favour dominant languages, while users speaking \u201csmaller\u201d languages have access to models that perform worse, or to no tools in their own language at all. Specialised national models are intended to close this gap. Polish LLMs, or Large Language Models, understand the Polish language, culture and context, and can also operate locally. They are developed primarily on the basis of Polish-language datasets and cultural heritage.<\/p>\n
\u201cSpecialisation in the Polish language also means that Bielik performs best in Polish tasks, legal contexts and linguistic nuances, for example when colloquial language appears,\u201d explains Sebastian Kondracki. \u201cIts popularity is so significant that we have reached 2.8 million downloads and, last month, we overtook Mistral. That is the French alternative, which has a large number of models, but in the compact model category Bielik recorded a higher monthly number of downloads.\u201d<\/p>\n
Bielik.AI is a family of European open-source language models. Since the beginning of 2026, it has supported 30 European languages and has formed part of the ecosystem of European social projects such as Eskadra Bielik, which aims to build the skills of AI agent creators across Poland, as well as Obywatel Bielik. The latter is a civic initiative focused on developing multimodal, inclusive and universally accessible artificial intelligence.<\/p>\n
\u201cWhat is very important is that Bielik no longer has to run on proprietary infrastructure only in Polish companies, but also in foreign ones. In June, we also want to release version 3.1, which will support 50 languages, including Chinese and Arabic,\u201d explains the CEO of Bielik.AI. \u201cWe are developing, and this is not the end. We organise many educational meetings and conferences, including the Bielik Summit, which will take place in November. A lot is happening, not only in the area of models, but also in building digital competences related to artificial intelligence.\u201d<\/p>\n
According to announcements, the Bielik Summit will take place on 26 November 2026. It will feature 120 speakers from Poland and abroad across four stages. Participants will be able to learn about more than 200 case studies of the use of AI, Bielik and compact models in business and public administration.<\/p>\n
Bielik.AI was developed to support Polish enterprises in data processing and to increase operational efficiency through the automation of customer service processes, data analysis and support for official documentation.<\/p>\n
\u201cThe main applications are in areas where compact models are strong. First of all, the Polish language. There are many applications in simplifying language. When we want to simplify a very bureaucratic text, for example for seniors, Bielik performs fantastically. In fact, it does so well that there is an application called Proste Pismo, developed on the basis of Bielik by researchers from Pozna\u0144,\u201d explains Sebastian Kondracki.<\/p>\n
Proste Pismo is a tool for simplifying the language layer of official letters. It was created in cooperation with linguists from Adam Mickiewicz University, the City of Pozna\u0144 Office and specialists from the Pozna\u0144 Supercomputing and Networking Center. The application supported, among other things, the simplification of content published on the website of the Polish Presidency of the Council of the EU, as well as other communication materials.<\/p>\n
\u201cSecondly, Bielik performs very well wherever there is a large volume of data and where we have to process huge amounts of information, for example in e-commerce, because we do not pay for every word or token. The more products we have to process, the higher the bills become \u2014 the so-called horror receipts. We download it free of charge, we have to invest in infrastructure, but we no longer pay for every processing operation. Bielik therefore works well wherever there is a lot of data, documentation or products,\u201d the expert notes. \u201cIt also applies to all kinds of sensitive data. Applications are being built on Bielik that support patient registration. It can process medical data without sending it to the cloud.\u201d<\/p>\n
Bielik.AI was created by the SpeakLeash Foundation in cooperation with researchers and the Academic Computer Centre Cyfronet AGH, which provided the computing resources of two supercomputers: Helios and Athena. The models are open source or operate under accessible licences, so anyone can download, use and develop them, adapting them to their own needs. Specialised models, for example medical ones, can be built on its basis. Bielik.AI has received patronage from the Ministry of Digital Affairs and is supported by companies.<\/p>\n
\u201cLet us remember that we are not a company, but a foundation. During the day, we work in business, and in the evenings we train Bielik. We do not have huge capital or investment funds behind us. We are maintained by sponsors and patrons. Of course, the biggest challenge is for Polish business to support us financially as well,\u201d emphasises Sebastian Kondracki. \u201cThe second thing is for people who start using artificial intelligence and Bielik to share their feedback with us, together with a justification. I keep repeating that we need a lot of thumbs down, and we need someone to tell us that in some areas we should still improve something. This is the most important knowledge we need in order for Bielik to become better and work efficiently in business.\u201d<\/p>\n
As he emphasises, operating in this model gives the creators the flexibility that is essential in AI development.<\/p>\n
\u201cWhen it comes to support from the public sector or the state, we receive it in the form of being able to train on Polish supercomputers. This is significant support. However, we are not applying for large subsidies or grants, because we would have to follow a fixed schedule, while in artificial intelligence everything happens so quickly that we have to move very flexibly,\u201d explains the CEO of Bielik.AI. \u201cIt is enough for a new scientific paper to appear somewhere in the world, and we already have to start training in those areas. If we had a grant with a two- or three-year schedule, it would be very difficult for us to maintain that flexibility. We need computing power and data from the state, and sometimes from citizens, but we will try to carry this out using our own resources or support from sponsors.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"
Bielik is one of Poland\u2019s AI language models which, unlike foreign alternatives, is specialised in the Polish language. Its key advantage for domestic users is that it can be run locally, on their own infrastructure, without the need to send data outside the organisation. The creators of Bielik have announced the release of a new […]<\/p>\n","protected":false},"author":1,"featured_media":5694,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2}},"categories":[11],"tags":[2974,2839,4656,2772,2690,4657,2753,4467,64,74],"class_list":["post-5693","post","type-post","status-publish","format-standard","has-post-thumbnail","category-technology","tag-artificial-intelligence","tag-content","tag-cyfronet-agh","tag-e-commerce","tag-gap","tag-helios","tag-lot","tag-open-source","tag-poland","tag-poznan"],"jetpack_publicize_connections":[],"_links":{"self":[{"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/posts\/5693","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/comments?post=5693"}],"version-history":[{"count":1,"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/posts\/5693\/revisions"}],"predecessor-version":[{"id":5695,"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/posts\/5693\/revisions\/5695"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/media\/5694"}],"wp:attachment":[{"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/media?parent=5693"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/categories?post=5693"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ceo.com.pl\/en\/wp-json\/wp\/v2\/tags?post=5693"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}