EMEA AI Model Builders
Explore and deploy sovereign AI models. Built to ensure data control, governance, and independence, these models are accelerated by NVIDIA’s AI inference platform and powered by NVIDIA-accelerated infrastructure.
Barcelona Supercomputing Center
ALIA is a collection of LLMs developed by the Barcelona Supercomputing Center (BSC). The collection includes Salamandra 2B, 7B, and ALIA 40B versions of both base and instruction fine-tuned models, pretrained on 9.37 trillion tokens across 35 European languages and code, prioritizing the official Spanish languages, including Castilian, Catalan, Basque, and Galician.
Explore
Explore sample applications to learn about different use cases for ALIA and Salamandra models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for LLMs with TensorRT-LLM. Learn how to set up and get started using ALIA and Salamandra in TRT-LLM.
Get started with the models for your development environment.
ALIA-40B Model
ALIA-40b is a highly multilingual model pre-trained from scratch that will come with its respective base and instruction-tuned variants.
Salamandra Model Card
Salamandra is a highly multilingual model pre-trained from scratch that comes in three different sizes — 2B, 7B and 40B parameters — with their respective base and instruction-tuned variants.
Black Forest Labs
FLUX is a family of advanced models engineered to transform imagination into precise and realistic images. From image generation to multi-stage image editing, creators can utilize natural language or multimedia context to control, transform, or refine visual content through an intuitive workflow. FP4-quantized checkpoints significantly reduce memory requirements, and NVIDIA TensorRT™ enhancements provide up to 2X faster inference.
Explore
Explore sample applications to learn about different use cases for FLUX models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for LLMs with TensorRT-LLM. Learn how to set up and get started using FLUX in TRT-LLM.
Get started with the models for your development environment.
FLUX.2 Model on Hugging Face
FLUX.2 by Black Forest Labs is a 32B parameter flow matching transformer model capable of generating and editing (multiple) images. The model is released under the FLUX.2-dev Non-Commercial License.
FLUX.1-dev Model by Black-Forest-Labs | NVIDIA NIM
FLUX.1-dev by Black Forest Labs is a cutting-edge generative AI model that creates high-quality images from simple text prompts.
Dicta
Dicta specializes in building high-performing, open language models for the Hebrew language, delivering robust performance, cultural awareness. The DictaBERT suite includes models fine-tuned for various difficult tasks in Hebrew using custom designed architectures, as well as providing powerful base models that can be fine-tuned for any task. DictaLM is a suite of generative large language models which set a new baseline for accuracy and fluency for on-prem Hebrew AI, leveraging advanced optimization through NVIDIA TensorRT-LLM and Dynamo for seamless production-scale deployment.
Explore
Explore sample applications to learn about different use cases for DICTA's models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for large language models (LLMs) with TensorRT-LLM. Learn how to set up and get started using TensorRT-LLM.
Get started with the models for your development environment.
Dicta-LM 3.0 Model Demo
Dicta-LM 3.0-thinking: Dicta-LM 3.0 is a powerful open-weight collection of LLMs, trained on extensive corpora of Hebrew and English texts.
DictaLM 2.0 - Instruct Chat Demo
Check out DictaLM 2.0 Instruct on Hugging Face to chat in Hebrew or English with an advanced AI language model fine-tuned for instruction-based conversations.
DictaBERT Collection
Collection of state-of-the-art language model for Hebrew, finetuned for various tasks,
ETH Zürich, EPFL, CSCS
Apertus is a collection of Switzerland’s first fully open, transparent, and multilingual foundation models featuring powerful Apertus 8B and 70B parameter models. Trained on over 15T tokens covering 1,000+ languages and approximately 40% non-English data, it includes underrepresented languages like Swiss German and Romansh. Collaboratively developed by ETH Zurich, the Swiss Federal Technology Institute of Lausanne (EPFL), and the Swiss National Supercomputing Centre (CSCS).
Explore
Explore sample applications to learn about different use cases for Apertus models.
Integrate
Get started with the right tools and frameworks for your AI model development environment.
Optimize
Deployment of the models is supported via the newest versions of LLAMA, MLX, SGLang, and vLLM.
Get started with the models for your development environment.
Apertus LLM Model Collection
Explore Apertus LLM on Hugging Face – Open Weights, Documentation, and Model Access.
Jetbrains
JetBrains is a global software company specializing in intelligent, productivity-enhancing tools for software developers and teams. Its open-source LLM, Mellum, optimizes code-related tasks such as code completion and intelligent code suggestions within developer environments. Following a Llama-style architecture, the transformer-based model has 4B parameters and an 8,192-token context window and is trained on 4T+ tokens spanning multiple programming languages.
Explore
Explore sample applications to learn about different use cases for Mellum models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for large language models (LLMs) with TensorRT-LLM. Learn how to set up and get started using TensorRT-LLM.
Get started with the models for your development environment.
AI Models from JetBrains: Powering Developer Productivity with Advanced Machine Learning Tools on Hugging Face
The JetBrains Hugging Face page hosts AI models and tools developed by JetBrains for improving developer productivity and coding with advanced machine learning techniques
Mistral
Mistral AI delivers industry-leading accuracy and efficiency for enterprise AI, such as their Mistral 3 family of open-source, multilingual, and multimodal models. Mistral’s models span frontier-level to compact models, enabling distributed intelligence across cloud and edge. Their models are optimized across NVIDIA supercomputing and edge platforms, empowering researchers and developers to experiment, customize, and accelerate AI innovation while democratizing access to frontier-class technologies.
Explore
Explore sample applications to learn about different use cases for Mistral models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for large language models (LLMs) with TensorRT-LLM. Learn how to set up and get started using TensorRT-LLM.
Get started with the models for your development environment.
Mistral Large 3
Mistral’s first mixture-of-experts model, Mistral Large 3 is trained with 41B active and 675B total parameters.
Ministral 3 14B
Ministral 3 14B Instruct 2512 FP8 is a powerful and efficient language model with vision capabilities, fine-tuned for instruction tasks.
Mistral Small 24B on NVIDIA Build
Mistral Small 3 (2501) sets a new benchmark in the "small" Large Language Models category below 70B, boasting 24 billion parameters and achieving state-of-the-art capabilities comparable to larger models.
SpeakLeash
Bielik is a family of open-source, lightweight LLMs (Apache-2.0) developed by the SpeakLeash Foundation and trained with ACK Cyfronet AGH. Designed for business applications and continuously improved through community-driven development, Bielik excels in document understanding, summarization, and knowledge extraction.
Explore
Explore sample applications to learn about different use cases for Bielik models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for large language models (LLMs) with TensorRT-LLM. Learn how to set up and get started using TensorRT-LLM.
Get started with the models for your development environment.
Bielik-11B on NVIDIA Build
Bielik-11B-v2.6-Instruct is a generative text model with 11 billion parameters, developed to process and understand the Polish language with high precision
Bielik-11B on Hugging Face
Bielik-11B-v2.6-Instruct is an 11-billion-parameter generative text model, fine-tuned for instruction following in Polish, developed by SpeakLeash and ACK Cyfronet AGH to deliver high-precision Polish language understanding and response generation
Technology Innovation Institute (TII)
The Falcon LLM series is a family of multilingual, open-source large language models developed by TII. Built for efficiency, long-context understanding, and strong reasoning, Falcon models contain multiple architectures designs, including Transformer, State Space Models (SSMs), hybrid Transformer-SSLMs, and support a wide range of global languages, including dedicated variants for Arabic and regional dialects. They excel in knowledge-intensive, STEM, and long-context tasks, making them highly versatile for diverse real-world applications.
Explore
Explore sample applications to learn about different use cases for Falcon models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for large language models (LLMs) with TensorRT-LLM. Learn how to set up and get started using TensorRT-LLM.
Get started with the models for your development environment.
Falcon3-7b-instruct Model by Tiiuae on NVIDIA NIM
The Falcon3-7B-Instruct is an open foundation model designed for state-of-the-art performance in reasoning, language understanding, instruction following, code generation, and mathematics. It supports long-context tasks with a token limit of up to 32K and multilingual capabilities in English, French, Spanish, and Portuguese.
Falcon H1R 7B Collection
Falcon H1R 7B is a compact, high-performance large language model designed to deliver world-class reasoning capabilities within a 7-billion-parameter architecture.
Falcon-H1 Family of Models and Falcon Arabic
Falcon-H1 Family of Hybrid-Head Language Models (Transformer-SSM), including 0.5B, 1.5B, 1.5B-Deep, 3B, 7B, and 34B (pretrained & instruction-tuned).
University College London (UCL)
UK-LLM is a sovereign AI initiative led by UCL, developed in partnership with NVIDIA and Bangor University to support speakers of minority UK languages like Welsh, Cornish, Irish, Scottish Gaelic, and Scots. It uses open-source techniques and datasets from NVIDIA Nemotron™ to enable advanced reasoning in English and Welsh, ensuring cultural and linguistic accuracy, for public services such as healthcare, education, and legal resources.
Explore
Explore sample applications to learn about different use cases for UK-LLM models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for large language models (LLMs) with TensorRT-LLM. Learn how to set up and get started using TensorRT-LLM.
Get started with the models for your development environment.
BritLLM on Hugging Face
BritLLM is a collection of multilingual language model training corpora and models, developed using machine-translated data from a single source language for broad pretraining applications
University of Ljubljana
The Generative Model for Slovene (GaMS) is a large language model family developed by researchers in the Centre for Language Resources and Technologies at the University of Ljubljana’s Faculty of Computer and Information Science. The family of LLMs is designed specifically for Slovene while also supporting other South Slavic languages and English.
Explore
Explore sample applications to learn about different use cases for GaMS models.
Integrate
Get started with the right tools and frameworks for your development environment.
Optimize
Optimize inference workloads for large language models (LLMs) with TensorRT-LLM. Learn how to set up and get started using TensorRT-LLM.
Get started with the models for your development environment.
Try the GaMS3 model
GaMS3-12B-Instruct represents the next generation of the GaMS (Generative Model for Slovene) model. The model is based on Google's Gemma 3 family and continually pretrained on Slovene, English, and some portion of Croatian, Serbian, and Bosnian corpora. The supervised fine-tuning phase was done on a combination of Slovene and English datasets.
GaMS-Instruct-Nemotron Collection
Next generation of GaMS models. Includes GaMS-Instruct models additionally improved using subset of NVIDIA Nemotron post training data.
More Resources
Ethical AI
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their supporting model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse. Please report security vulnerabilities or NVIDIA AI Concerns here.