Changelog

Follow our journey as we build the future of on-device AI training.

Late 2024

Launched Embasis Group v1.0 with full support for MLX quantization on iOS devices.

Core Features

  • 4-bit quantization support
  • On-device fine-tuning for Llama 3
  • Metal performance shaders optimization

Mid 2024

Beta testing began with select partners. Focused on optimizing memory usage for iPhone 15 Pro.

✅ Reduced memory footprint by 40%
✅ Implemented LoRA adapters
✅ Added support for custom datasets

Early 2024

Initial prototype and proof of concept.

We successfully ran a 7B parameter model locally on an iPad Pro for the first time.