Open Superintelligence Lab

Making AI development accessible with scalable superintelligence research

Our Vision

Any company or person (even with no technical experience) should be able to download this repository and run it on their GPU setup - from 1 GPU to 1 million GPUs. The system will be able to automatically detect your hardware configuration, tune hyperparameters for optimal performance, and run the best possible training with or without manual configuration from your side.

🚀

Auto-Scaling

Seamlessly scale from single GPU to massive distributed clusters

Auto-Tuning

Intelligent hyperparameter optimization for your hardware

🔧

Zero-Config

Works out of the box with automatic hardware detection

Blueberry LLM 🫐

Our flagship Mixture of Experts (MoE) language model implementation. Clone, install dependencies, and train your own language model with a single command. Perfect for researchers and developers looking to experiment with cutting-edge AI architectures.