Startup Mindbeam AI releases Litespark-Inference, an open-source framework using ternary models to run LLMs on CPUs up to 96 times faster, significantly reducing memory use.
