According to Tom's Hardware, Cyanogen co-founder Kirt McMaster announced Kora today, a new Linux desktop operating system designed to split AI workloads across CPU, GPU, NPU resources, and other machines connected to the same local network. The OS, built by Kora Systems Inc., features a proprietary Cortex hybrid inference engine optimized for running local AI models by default and cloud models when needed.
Kora currently supports the Snapdragon X2 Elite's Hexagon NPU, which runs Google's Gemma 4 26B model at 25 tokens per second for decoding. The company recommends 64GB of RAM to access better models and projects that approximately 70 percent of model calls will run locally in 2027. Support for Intel and AMD NPUs is planned, though AMD's implementation will likely be limited to embedding and speech-recognition models. The operating system is expected to arrive on PCs in 2027, with the enterprise variant Humain OS becoming commercially available first.