The LiquidAI LFM2.5-230M model is now running locally in the browser using custom WebGPU kernels. These specialized kernels were originally developed by Fable 5 prior to its shutdown and Opus 4.8. The demonstration was recorded on an M4 Max device, achieving a generation speed of 1,400 tokens per second. All processing occurs entirely within the user's browser environment without external server dependencies. A GGUF version of the model is available for download on Hugging Face alongside the standard checkpoint. Users can interact with the live demo hosted by the webml-community on Hugging Face Spaces.