Italia

AI model · Language model (LLM) · Text → Text by Domyn

The open-source foundational LLM for sovereign financial intelligence.

Italia is an open-source foundational LLM designed for the financial services industry, enabling institutions to deploy sovereign AI on their own infrastructure. It was trained on the Leonardo supercomputer using NVIDIA technology to support mission-critical financial reasoning.

Model specs

The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.

Languages
Italian, English

Category: Generative AI

License
Open-source
Released
July 2024
Target Industry
Financial Services (FSI)
Model Type
Foundational Large Language Model
License Type
Open-source
Training Infrastructure
Leonardo Supercomputer (CINECA)
  • Open-source foundational LLM for FSI
  • Designed for sovereign on-premise deployment
  • Trained on the Leonardo supercomputer
  • Built with NVIDIA NeMo and Megatron-Core

Overview

Italia is an open-source foundational large language model developed by Domyn specifically for the financial services industry (FSI). Launched in July 2024, it serves as a specialized foundation for financial reasoning, compliance, and domain-specific enterprise tasks. The model is designed to provide a sovereign AI solution, allowing financial institutions to run the model on their own infrastructure to maintain full control over sensitive data. It was built using advanced training pipelines, including NVIDIA NeMo and Megatron-Core, and was trained on the Leonardo supercomputer. By offering an open-source foundational model, Domyn enables regulated organizations to achieve technological independence from external providers.

Financial Domain Expertise

  • FSI-Specific Design: The model is specifically designed to be used by organizations in the financial services industry for domain-specific tasks.
  • Sovereign Infrastructure: It provides a sovereign AI solution that allows financial institutions to run models on their own infrastructure to keep sensitive data in-house.
  • Open-Source Foundation: The model is released as an open-source foundational large language model to support global AI ownership.

Advanced Training Architecture

  • NVIDIA NeMo Integration: The model was built using the NVIDIA NeMo framework for large-scale language model development.
  • Megatron-Core Optimization: Training utilized Megatron-Core to optimize performance for large-scale foundational models.

Use cases

  • Financial institutions run the model on their own infrastructure to keep sensitive customer data in-house.
  • FSI organizations use the specialized foundation to perform complex financial reasoning and compliance tasks.
  • Regulated enterprises leverage the open-source weights to build sovereign AI stacks without external dependencies.

Specifications

License
Open-source
Released
July 2024
Target Industry
Financial Services (FSI)
Model Type
Foundational Large Language Model
License Type
Open-source
Training Infrastructure
Leonardo Supercomputer (CINECA)
Training Frameworks
NVIDIA NeMo, Megatron-Core
Deployment Mode
On-premise / Private Infrastructure

FAQ

What is Italia?

Italia is an open-source foundational large language model designed specifically for the financial services industry to enable sovereign AI deployment.

How does Italia handle data privacy?

The model was developed to allow financial institutions to run AI on their own infrastructure, ensuring sensitive data remains in-house and under their full control.

What technology was used to train Italia?

Italia was trained using NVIDIA NeMo and Megatron-Core on the Leonardo supercomputer.

Who is Italia for?

It is intended for organizations in the financial services industry that require specialized reasoning and compliance capabilities.

Added by AntSpace from public sources (domyn.com).

Related

More AI models by Domyn