Yeco Flash Mini-IT: The Italian Micro-LLM That Challenges Silicon Valley at Zero Cost

In a global AI market dominated by multimillion-dollar supercomputers in San Francisco and Beijing, the Italian startup YecoAI — led by 18-year-old Marco Nasi from Piedmont — has released an unprecedented engineering provocation: Yeco Flash Mini-IT.

This is not another commercial "wrapper" — a graphical interface borrowing its brain via API from American servers. Yeco Flash Mini-IT is a Proof of Concept of a native proprietary LLM, built and trained entirely in-house.

The numbers of an economic and energy miracle

• 118 million parameters: an ultra-light size conceived for fast logical reasoning in Italian
• Trained for under €15: an absolute record of code optimization that slashes academic and industrial development costs
• Local CPU execution: no third-party cloud needed. Runs natively on very old consumer hardware at ~200 tokens per second

Demolishing giants on logic benchmarks

The real strength of Yeco Flash Mini-IT lies in official benchmarks. On Italian computational logic tests, this micro-model outperforms solutions five times larger (like Minerva-350M on BoolQ: 62.0 vs 60.7, measured on ITA-Bench), proving that hyper-specialized code and clean proprietary datasets matter far more than raw power.

Sovereign stack, together with Yeco-PII

Paired with Yeco-PII — our 278M-parameter shield model that anonymizes sensitive data beating OpenAI in six European languages with an F1 of 0.966 vs 0.769 — Yeco Flash Mini-IT is the first true infrastructure-grade, enterprise alternative for companies demanding total digital sovereignty, full GDPR and EU AI Act compliance, and inference costs near zero.

Sovereign technology is no longer a philosophical concept for institutional debate: it is an efficient, lightweight, independent silicon engine. And it has only 118 million parameters.

The model is Apache 2.0 licensed and subject to rework as the series evolves. Read the full model card on its research page.