---
title: "Yeco Flash Mini-IT v1 · YecoAI Research"
description: "118M-parameter Italian micro-LLM trained for under €15, running at ~200 tok/s on CPU. Measured ITA-Bench results. Apache 2.0."
url: "https://yecoai.com/en/research/yeco-flash-mini-it"
locale: "en"
published: "2026-09-29"
updated: "2026-09-29"
category: "Paper"
---

# Yeco Flash Mini-IT v1

118M-parameter Italian micro-LLM, trained for under €15, ~200 tok/s on CPU. Beats 5× larger models on Italian logic benchmarks.

*29 Sep 2026 · Paper*

## Abstract

Yeco Flash Mini-IT is a proof-of-concept proprietary Italian LLM built and trained entirely in-house. While frontier labs burn billions on trillion-parameter models, this 118M-parameter architecture (118.1M measured) shows that hyper-specialized code and clean proprietary datasets matter more than brute force.

## Efficiency record

- 118M parameters (GPT-2 decoder-only + Context Gating, Pre-LN, weight tying)
- Trained for under €15 total (pretraining \~$8.75 on H100 + SFT \~$4)
- Runs locally on CPU: \~200 tokens/second on old consumer hardware
- Custom Italian BPE 32k tokenizer, fertility 1.53 tok/word (−34% vs GPT-2)

## Benchmarks (measured, ITA-Bench)

Scores are measured on official ITA-Bench datasets (SapienzaNLP), log-likelihood per choice, N=200/task, seed 42 — not estimated.

- ARC-Challenge: 24.0% (Minerva-350M paper: 24.6%)
- HellaSwag: 27.5% (Minerva: 32.6%)
- BoolQ: 62.0% — beats Minerva-350M (60.7%)
- TruthfulQA: 34.5% · GSM8K: 32.0% · Average: 36.0%

Honest reading: on multiple-choice benchmarks V2 is within statistical noise (±3.5 points/task) of the pre-SFT checkpoint. V2's value is behavioral: reliable question→answer→stop format, echo eliminated, structured chat answers.

## License

Apache 2.0. Subject to rework: the model is under active iteration and numbers may change with future releases.

```json
{
  "@context": "https://schema.org",
  "@id": "https://yecoai.com/en/research/yeco-flash-mini-it#page",
  "url": "https://yecoai.com/en/research/yeco-flash-mini-it",
  "name": "Yeco Flash Mini-IT v1 · YecoAI Research",
  "description": "118M-parameter Italian micro-LLM trained for under €15, running at ~200 tok/s on CPU. Measured ITA-Bench results. Apache 2.0.",
  "inLanguage": "en",
  "isPartOf": {
    "@id": "https://yecoai.com/#website"
  },
  "@type": "ScholarlyArticle",
  "headline": "Yeco Flash Mini-IT v1",
  "mainEntityOfPage": "https://yecoai.com/en/research/yeco-flash-mini-it",
  "datePublished": "2026-09-29T00:00:00.000Z",
  "dateModified": "2026-09-29T16:17:32.482Z",
  "articleSection": "Paper",
  "keywords": "YecoAI Research, Yeco Flash Mini-IT, micro-LLM, ITA-Bench, Italian model, efficient training",
  "author": {
    "@id": "https://yecoai.com/#organization"
  },
  "publisher": {
    "@id": "https://yecoai.com/#organization"
  }
}
```

```json
{
  "@context": "https://schema.org",
  "@type": "BreadcrumbList",
  "itemListElement": [
    {
      "@type": "ListItem",
      "position": 1,
      "name": "YecoAI home",
      "item": "https://yecoai.com/en"
    },
    {
      "@type": "ListItem",
      "position": 2,
      "name": "Research",
      "item": "https://yecoai.com/en/research"
    },
    {
      "@type": "ListItem",
      "position": 3,
      "name": "Yeco Flash Mini-IT v1",
      "item": "https://yecoai.com/en/research/yeco-flash-mini-it"
    }
  ]
}
```