Welcome to ThingsAI! 🤗

Building efficient, bilingual AI models that run anywhere. 🇮🇹 🇬🇧

🤖 Our Models

Quark-135M

A lightweight bilingual (Italian + English) language model with 135M parameters. Features GQA, SwiGLU, RMSNorm, and RoPE. Trained on 50B+ tokens of curated Italian and English data.

Quark-270M (training now)

Our most powerful small model — 270M parameters with 32 layers, 768 hidden dimensions, and 65K vocabulary. Currently in active training on 10B+ tokens with planned scaling to 135B tokens.

Quark-Mod

A multi-label moderation model covering 9 categories for safe AI deployment: toxic, severe_toxic, obscene, threat, insult, identity_hate, cyberbullying, hate_speech, offensive.

🎯 What We Focus On

• ⚡ Small, efficient architectures — GQA, weight tying, deep-thin design
• 🌍 Bilingual training — Italian + English from scratch
• 🔓 Open-source everything — weights, code, datasets
• 💻 Real-world deployment — runs on consumer hardware

📂 Resources

_{Made with ❤️ by ThingsAI · Website · GitHub}