Yeco Flash Mini-IT: an Italian Micro-LLM Trained for Under €15
118M parameters, trained for under €15, ~200 tok/s on a consumer CPU. Yeco Flash Mini-IT is our proof of concept that sovereignty beats scale.
Twenty days of redesign, from September 10 to September 30, 2026. The new YecoAI brand and official website are live, designed by Kevin Paratore of IdraArts.
Marco Nasi
Founder & CEO
Twenty days of work are behind the site you are reading this on: today, September 30, 2026, the new YecoAI brand identity and official website are live. This is the story of the redesign, and of the people who made it possible.
The work started on September 10 and shipped today. In between: a new visual identity, a rebuilt design system, and a website designed to say exactly what we do, with numbers instead of adjectives.
YecoAI builds AI systems that work: privacy models that beat far larger alternatives, support automation for hosting infrastructure, gaming platforms scaled to tens of thousands of users. The previous site did not tell that story clearly. The new one does: every product page leads with measurable results, every claim is backed by benchmarks, and the design keeps the interface out of the way of the content.
The redesign also brought structure: bilingual content in English and Italian, a blog and research section with proper feeds, llms.txt and Markdown mirrors of every page, and an investors page with a full pitch deck and a dedicated request form.
The UI and UX design of the new site is the work of Kevin Paratore, freelance designer behind IdraArts. Kevin designed the brand identity, the layout system and every page of the interface, turning our brief into something coherent and precise. He is now listed on our team page as an external collaborator: he is a freelancer, not an employee, and the credit is well deserved.
Grazie Kevin, e grazie a IdraArts per il lavoro svolto in queste tre settimane.
The next milestones are already on the timeline: the SRL incorporation by Q4 2026, DeskNexo beta, and the Yeco LLM Series releases. If you want to talk about them, the contact page is one click away.
118M parameters, trained for under €15, ~200 tok/s on a consumer CPU. Yeco Flash Mini-IT is our proof of concept that sovereignty beats scale.
Ender-1 is our proprietary 3B model trained on EnderDevelopment data (free users, policy-compliant). In beta: staged rollout beginning in shadow mode.
278M-parameter neural model for PII detection and anonymization. Native Italian, 6 European languages, average F1 0.966.