Google open-sources ML Drift GPU engine for on-device AI inference

Google released ML Drift, an Apache 2.0 cross-platform GPU compute engine for on-device AI inference supporting OpenGL ES, OpenCL, Metal and WebGPU. It powers LiteRT, cuts memory overhead by 12% versus llama.cpp and MLX LM, and delivers up to 40% lower frame latency in YouTube Shorts.

• Supports OpenGL ES, OpenCL, Metal and WebGPU under Apache 2.0
• 12% lower memory overhead than llama.cpp and MLX LM
• Up to 40% lower frame latency in YouTube Shorts on Android and iOS
• Up to 30% speedup for AI features in Lightroom and Photoshop