Mario-1: an Italian language model that runs on a CPU
A 0.8-billion-parameter language model built for Italian. It runs on a CPU without a GPU, recognises trick questions, admits what it does not know and does reliable arithmetic.


Italian and European models, trained for one precise task and used in our products: DeskNexo for hosting support, EnderDevelopment for gaming communities, Yeco-PII for data privacy.
We build AI that solves vertical problems end-to-end: proprietary models, deterministic safety layers, production-grade evaluation.

Every product is built on measurable research: published benchmarks, honest model cards, reproducible evaluations.

Designed for GDPR and the EU AI Act. Our models run on your infrastructure: no third-party API brains, no data leaving your systems.

Compact models on commodity hardware: PII anonymization in milliseconds, support tickets resolved in under 30 seconds.

Our new model that finds and hides personal data in text. More accurate than the previous version and better than much larger models, especially in Italian. Every release is digitally signed, so you can verify it is genuine.

118.1M-parameter Italian micro-LLM, trained for under €15, ~200 tok/s on CPU. Beats Minerva-350M (3× its size) on BoolQ-IT.

Our AI model for EnderDevelopment, a fine-tune of Qwen2.5-Coder-1.5B (Apache 2.0). It is used only inside the product and is not publicly available.

DeskNexo is the AI support agent for hosting providers: it reads WHMCS tickets, diagnoses VMs on Proxmox and VirtFusion, resolves the issue, and writes a human reply to your customer, alone, in under 30 seconds, 24/7. Every action is snapshot-protected and dual-LLM verified.

EnderDevelopment is the AI platform for gaming communities: natural language in, working Minecraft plugins (Spigot/Paper), FiveM scripts, Discord bots and server configs out. Free plan, credit-based pricing from €8/month, and the real-world training ground for Ender-1.

Yeco LLM Series is the family of proprietary Italian AI models from YecoAI: Yeco-PII v1.0 (278M, privacy) and Yeco Flash Mini-IT v1 (118M, trained for under €15). Ender-1, our fine-tune of Qwen2.5-Coder-1.5B, is a separate product model. Sovereign, efficient, compliance-ready, this page updates as the series grows.

Over 30,000 registered users on EnderDevelopment. We rebuilt the platform from the ground up into a fully modular experience, engineered for the next million users and managed with total control.

YecoAI is accepted into Anthropic's Claude Startups program: one year of Claude Team free and over $1,000 in Claude API credits. For us, that support is huge.

Two awards in one weekend: YecoAI takes home the Premio Aulab and finishes third at DigithON 2026. A heartfelt thank you to DigithON, Aulab, and everyone who believed in us.

Twenty days of redesign, from August 10 to September 30, 2026. The new YecoAI brand and official website are live, designed by Kevin Paratore of IdraArts.

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 a fine-tune of Qwen2.5-Coder-1.5B (Apache 2.0) trained on EnderDevelopment data. In beta, with a staged rollout in shadow mode.
Institutional programs and collaborations that support our work.
The Italian AI that solves problems, not demos? Get in touch
A sovereign AI research company based in Mondovì, Italy. We research small models trained for one task, and we put that research into our own products: DeskNexo, EnderDevelopment and Yeco-PII.
Think of Google. It started as a research lab, then launched products like AdSense to fund itself. We work the same way: sovereign AI research is the core, and we apply it in our own products. Those products fund us, so we don't depend only on outside capital.
Models designed, trained and evaluated by our team in Italy, built to run on infrastructure the customer controls, instead of renting a general-purpose API from abroad. Sensitive data can stay inside the customer's perimeter.
They are in use. DeskNexo automates WHMCS hosting support end-to-end. EnderDevelopment has 30K+ registered users. Yeco-PII v1.0 is released and, in our published evaluation, outperforms the OpenAI Privacy Filter.
Our products do, with public support from Invitalia and the AWS Startup Program. The goal is to grow without depending only on outside capital.