AIGENCY Public Announcement

Public Announcement Regarding Our National Artificial Intelligence Initiative

Dear public, esteemed technology and AI community,

In recent days, various technical questions, evaluations, and criticisms have surfaced on social media and digital platforms regarding our national artificial intelligence platform, AIGENCY. This interest reflects both the impact our project has achieved and the growing awareness of artificial intelligence in our country. In this context, we would like to fulfill our responsibility to inform the public directly and transparently about the project.

1. AIGENCY’s Architectural Structure and Technological Approach

At its core, AIGENCY is based on the open-source, multilingual LLAMA 3 architecture with 70B parameters, which has been extended with our own uniquely developed 140B parameter Turkish layer and auxiliary learning modules, forming a hybrid model. The current total system parameters have reached 210 billion.

This expansion is not merely a language extension but also focuses on:

• Turkish contextual understanding capability,
• Conceptual relations and cultural semantic accuracy,
• Safe content generation,
• Filtering unnecessary directions and brand/advertisement content.

These elements have been developed into an original, controlled, and user-centric structure. These efforts go far beyond a classical fine-tuning process. The primary goal has been to create an AI that can work directly with user experience and local data integrity.

2. Performance Tests and Competency Criteria

Throughout its development, AIGENCY has undergone various benchmark tests, including:

• MMLU (90.1% success rate) – Strong contextual analysis in multidisciplinary academic questions,
• GPQA (50.4% success rate) – Complex logical inference and data analysis,
• HumanEval (91.2% accuracy) – High precision in software and algorithm generation.

Each of these universal measurement tools demonstrates that AIGENCY is a competitive solution not only in Turkish but also in terms of its technical infrastructure.

3. Misleading Comments and Technical Misconceptions

In some circles, evaluations disconnected from the technical context have been made regarding our model’s system architecture. For example, instances where a user prompt is mistaken for a system prompt or misinterpretations of temporary test visuals during software updates. Such interpretations do not reflect the quality of the project and are far from technical reality.

We also want to point out that we have been subjected to intense cyberattacks over the past three days. The simultaneous appearance of certain social media comments with similar tone and content during these attacks is noteworthy. The data logs we have obtained are also leading us to initiate a legal process in this regard. Nonetheless, we continue to evaluate all constructive feedback with care, as it contributes to our roadmap.

4. The Value Represented by the Terms "First" and "National"

Some comments question the term "first Turkish LLM." What must be emphasized here is that we are not claiming to have developed a model entirely from scratch. Rather, we created an AI platform developed by a local team using local resources, prioritizing the Turkish language. The term “national” here refers to the fact that our engineering, vision, and data structure are based in Turkey. This is a declaration of intent more than a technical claim.

It should be remembered that the true value of digital transformation lies not in producing a model from scratch, but in adapting existing knowledge infrastructures to local needs.

5. Transparency to the Public and Future Steps

As AIGENCY, as we have done so far, we will continue to publish:

• All technical documents,
• API usage documentation,
• Model updates,
• Experience reports

via our official website and press channels. Each statement will be made on the basis of technical accuracy and user security.

AIGENCY is more than just a software product. Even taking the first step in AI development in our country is valuable. We have taken that step. We are open to all kinds of contributions. However, we also declare that we will not be a target of unethical comments not based on facts.

We thank the thousands of users who trust us, all our teammates who contributed, and everyone who offers constructive criticism.

Our journey is long. Our belief is strong.

Respectfully,

The AIGENCY Team

eCloud Software Technologies