🚀 VeroLM

Fast, efficient, high-quality models. Mostly open source, free for everyone.

VeroLM is an independent AI lab training language models at every scale, from tiny edge-friendly models to large ones. Most of our work is released as open source and is freely accessible to everyone.

💡 What we believe

• Quality first. We build models that genuinely perform, not just models that exist.
• Speed and efficiency. Every model should be as fast and as optimized as we can make it, so it runs on the hardware people actually have.
• Open source by default. Most of what we train is released openly and for free. Good AI shouldn't sit behind a paywall.
• Experimentation. We constantly try new and rarely explored ideas in architecture, training, and optimization, and we share what we learn.

📦 What you'll find here

• Models of many sizes, from small and lightweight to larger and more capable
• Experimental releases built around new and uncommon techniques
• Optimized variants focused on inference speed and low resource usage

🤝 Stay in touch

New releases land regularly. Follow the organization to get notified, and feel free to open a discussion on any model page with feedback, ideas, or questions.

Built with curiosity, released for everyone.