Slovak Language AI Hub
The central hub for AI and language technologies for Slovak.
Slovak Language AI Hub (SLAIH) is an initiative aimed at supporting the development, sharing, and use of language technologies for Slovak. We connect people, teams, and organizations that are advancing natural language processing in Slovakia.
Discover datasets, tools, benchmarks, and community in one place.
The best models for Slovak right now
- Speech to textlower is better
- 1Microsoft MAI-Transcribe-26.54%
- 2Google Chirp 38.14%
- 3ElevenLabs Scribe v29.15%
- Generative modelshigher is better
- 1Gemini 3.1 Pro72.65
- 2GPT-5.571.02
- 3Claude Opus 4.668.13
- Text understandinghigher is better
- 1mDeBERTaV385.17
- 2FERNET-CC84.27
- 3mmBERT84.17
- Text embeddingshigher is better
- 1multilingual-e5-large-instruct77.49
- 2gemini-embedding-00177.23
- 3jina-embeddings-v375.10
Latest news
View all →NLP Summer School 2026: Three Days Exploring Language, AI and the Future of Intelligent Systems
EventMore than a third of the participants came from abroad, bringing students, PhD candidates, researchers, and AI practitioners together in Bratislava for three days of discussions about the current state and future direction of NLP. NLP Summer School 2026 explored large language models, AI agents, multimodal AI, Slovak language technologies, and the relationship between artificial intelligence and people through lectures, discussions, and practical perspectives.
Published September 24, 20265 min readHow (not) to read benchmarks and leaderboards
Know-howIt's easy to rely on a single number from a leaderboard when choosing a model for your use case. But that number can hide far more than it reveals. It's important to also look at how informative the benchmarks used actually are, and what capabilities they evaluate.
Published August 21, 202610 min readWhat's New in the SLAIH Asset Catalog? Five Resources Worth Exploring
AnnouncementOne of the main goals of SLAIH is to make it easier to discover high-quality resources for Slovak language AI. In this edition, we'd like to highlight a few recent additions that we believe deserve your attention. Whether you're building an LLM application, training an ASR system, or experimenting with multilingual models, these resources are well worth exploring.
Published July 30, 20265 min read
Recently added resources
All resources (31)- DatasetTrePal
TrePal is an open Slovak speech dataset from plenary sessions of the National Council of the Slovak Republic, for training and evaluating automatic speech recognition models.
- Streams s.r.o.
Transcribe speechCC BY 4.0 - Transcribe speechCC BY 4.0
- Transcribe speechCC BY 4.0
- Transcribe speechApache 2.0
- Transcribe speechApache 2.0
By task
All tasksTranscribe speech10 resources
Transcribe Slovak speech from meetings, interviews, lectures, and phone calls.
Who works on it
- DatasetTrePalStreams s.r.o.
- ModelParakeet TDT 0.6b v3 SKNLP KInIT, KInIT
- ModelCanary 1B v2 SKNLP KInIT, KInIT
Turn text into speech3 resources
Models that read Slovak text aloud.
Who works on it
So far, all authors are outside the SLAIH community.
- ModelOmniVoiceXiaomi Corp.
- ModelSupertonic-3Supertone
- ModelPiperOpen Home Foundation
Search by meaning5 resources
Embedding models and datasets for searching, comparing, and clustering texts by meaning.
Who works on it
- ModelBGE-M3University of Science and Technology of China, BAAI
- DatasetSkMTEBSlovakNLP Community
- ModelE52 sizesLRKT FEI TUKE, NLP KInIT, NaiveNeuron FMPI CU
Understand text3 resources
Models and datasets for part-of-speech tagging, named entity recognition, natural language inference, and question answering.
Who works on it
- DatasetANLI SKNLP KInIT, DFKI
- ModelSlovakBERT NLP KInIT, Gerulata
- DatasetskLEPLRKT FEI TUKE, NLP KInIT, NaiveNeuron FMPI CU
Write in Slovak3 resources
Language models that generate Slovak text and the corpora used to train them.
Who works on it
- ModelMiniMax-M3MiniMax
- DatasetFineWeb2 SlovakNLP KInIT
- ModelQwen3 14B SKJÚĽŠ SAV, LIMAD UUI TUKE, CSPV SAV
Check text and models3 resources
Data for detecting machine-generated text, text errors, and gender stereotypes in models.
Who works on it
- DatasetMultiSocialWUDAP KInIT
- DatasetGESTNLP KInIT, WUDAP KInIT, Ethics KInIT
- DatasetChibyJÚĽŠ SAV
More resources4 resources
Speech databases, dictionaries, and lists that are not specific to one task.
Who works on it
- DatasetALOIS-DBSpeech UI SAV
- DatasetEWA-DBSpeech UI SAV
- ToolSlovak ResourcesSlovakNLP Community
Get involved
People from universities, research institutes, and companies contribute to SLAIH. You can contribute too.
Add a model or dataset
Send a Hugging Face or GitHub link, license details, and a few sentences about what the resource does. We will list it under the right task.
Submit a resourceGet your model evaluated
We evaluate speech recognition, embedding, and text understanding models on Slovak data. Send us a link to your model and see it on the leaderboard alongside the others.
Submit a modelIntroduce your team
Working on language technology for Slovak? We will add your team and its resources to the community.
Add a teamContribute a news article
Have new results, an upcoming event, or a tutorial to share? Write about it, and we will publish your article.
Suggest an article