Turkish Legal AI Project "TurkLeLM" Secures Supercomputing Grant to Revolutionize the Sector

• Our TurkLeLM project has secured 18,000 GPU hours on Europe’s MareNostrum 5 supercomputer through the TRUBA program to develop cutting-edge AI models tailored specifically for the Turkish legal system.

• Built from a vast dataset including 100+ billion tokens of Turkish legal documents, TurkLeLM aims to overcome limitations of global AI models by delivering unmatched accuracy and deep understanding of Turkish law.

• Committed to open-source development, we are creating public AI infrastructure to empower Türkiye’s legal and tech sectors while strengthening national technological sovereignty.

We are proud to announce that our TurkLeLM (Turkish Legal Language Models) project has been awarded 18,000 GPU hours on Europe’s premier MareNostrum 5 supercomputer through the prestigious TRUBA program. This grant provides us with the computational power needed to develop the most advanced AI models ever created specifically for the Turkish legal system, with the goal of transforming legal practice across Türkiye.

Our project addresses a critical gap in the market. While global AI models are powerful, they consistently struggle to grasp the unique terminology, procedural nuances, and contextual subtleties of Turkish law—challenges that have limited innovation in Türkiye’s legal sector. To overcome this, we are building TurkLeLM from the ground up using a massive, carefully curated dataset of 300 billion tokens, including an unprecedented 120 billion tokens sourced from Turkish legal documents. This ensures our models achieve unparalleled accuracy and domain-specific understanding.

A Sophisticated Two-Pronged Technical Strategy

We are advancing TurkLeLM through a dual-architecture approach designed for both immediate impact and long-term innovation:

• Our primary focus is on developing state-of-the-art legal embedding models. These specialized models act as hyper-intelligent indexes for legal texts, enabling a deep understanding of meaning and context beyond simple keyword matching. This technology powers superior semantic search tools that allow legal professionals to find relevant case law and statutes with unprecedented precision.

• At the same time, we are progressing on fully generative Turkish legal Large Language Models (LLMs). By continually pre-training powerful open-source models, we aim to build AI capable of drafting legal documents, summarizing complex cases, and delivering sophisticated legal analysis.

Our Open-Source Commitment for National Impact

Central to our mission is a commitment to open-source development. Unlike proprietary corporate AI, we are creating a public software infrastructure to empower Türkiye’s entire legal and tech ecosystem—from startups and legal-tech innovators to established law firms and academic institutions.

This initiative not only advances legal technology but also strengthens Türkiye’s AI sovereignty by fostering indigenous AI capabilities tailored to local needs. By reducing dependence on foreign AI solutions, we are helping build a new generation of Turkish AI expertise.

We expect to release the first open-source embedding models by early 2026, followed shortly by the full generative LLMs—marking a new era for legal technology in Türkiye.

About TRUBA Program

The TRUBA (Turkish National Academic Network and Information Center) program fosters high-impact research by granting access to world-class computational resources through strategic partnerships with leading European supercomputing centers. Designed to accelerate research excellence and drive technological innovation in Türkiye, TRUBA enables projects to leverage cutting-edge infrastructure. Recognition by the TRUBA program grants privileged access to the MareNostrum 5 supercomputer—one of Europe’s most powerful systems—housed at the Barcelona Supercomputing Center.

Stay Connected and Join Us on This Journey

Follow our progress as we develop TurkLeLM and redefine legal technology in Türkiye. Subscribe to our blog and connect with us on LinkedIn to receive the latest updates, research breakthroughs, and announcements about upcoming open-source releases. Together, we’re building the future of AI-powered legal services—don’t miss a moment of this transformative journey.