Powering the Next Generation of Greek and European AI

Layer42Labs is led by the researchers behind Meltemi —the first open LLM for the Greek language— and its state-of-the-art successor, Krikri-8B.

We operationalize open-source research for the enterprise. We develop state-of-the-art, multilingual and multimodal AI that ensures the inclusion of the Greek language, offering companies optimized deployment and fine-tuning services that surpass the capabilities of standard models.

Layer42Labs Logo

From Research to Industry

Layer42Labs is led by the ILSP/Athena RC scientific team that architected and trained the first state-of-the-art Greek models.

Sokratis Sofianopoulos

Sokratis Sofianopoulos

CEO, Systems and Software Design

Dimitris Roussis

Dimitris Roussis

Data Synthesis and Processing

Leon Voukoutis

Leon Voukoutis

Model Training, Optimization

George Paraskevopoulos

George Paraskevopoulos

System Design, Multimodal ML

Prokopis Prokopidis

Prokopis Prokopidis

NLP Applications, Evaluation

Vassilis Katsouros

Vassilis Katsouros

Scientific Lead

Strategic Parnerships

Layer42Labs collaborates with major technology players to access cutting-edge resources, expertise, and infrastructure that push the frontiers of AI

GRNET
Amazon
Google
IBM

Technical Capabilities

Layer42Labs specializes in deploying and fine-tuning open LLMs for enterprise environments.

Efficient Tokenization

Our models use custom tokenizers with up to ~21k Greek tokens, reducing fertility to 1.65 and lowering inference costs by ~40%.

Instruction & Alignment

Leveraging our synthetic data pipeline and state-of-the-art preference optimization techniques, chat versions of our models are designed to follow instructions safely and effectively.

Modern Greek Focus

Optimized for Modern Greek business applications, with additional support for Polytonic and Ancient Greek.

Code & Reasoning

Maintaining strong performance in English code generation and math reasoning, essential for complex technical workflows.

Enterprise Integration

Step 1: Audit

Our team assesses your data readiness and infrastructure needs.

Step 2: Deployment

We deploy our latest open model (Krikri-8B) within your secure private cloud or on-premise servers.

Step 3: Adaptation

We fine-tune the base model on your proprietary data to achieve maximum domain accuracy.

Live Production

AI Pipeline

Based on Open Peer-Reviewed Research

Our research, recently presented at EMNLP 2025 and arxiv, demonstrates that specialized training significantly outperforms generic multilingual models.

Scientific Breakthrough

The Krikri-8B model established a new standard for Greek NLP.

VoxKrikri

We introduced VoxKrikri, the first Greek speech LLM, achieving state-of-the-art results for Automatic Speech Recognition in Greek and providing an average ~20% relative improvement across benchmarks.

Open Benchmarks

We have contributed more than 20 open benchmarks to evaluate model performance.

Contact Us

info@layer42labs.eu