Mindbeam Claims 96x AI Speed Boost on CPUs

Startup Mindbeam AI has released Litespark-Inference, an open-source framework enabling ternary large language models to run on standard CPUs from Apple, Intel, AMD, and Arm. The company claims throughput improvements of up to 96-fold over standard PyTorch implementations, with memory usage cut by over 80%. Rather than replacing GPUs, Mindbeam positions CPUs as complementary accelerators in AI inference pipelines. Benchmarks show an Apple M5 chip achieving nearly 40 tokens per second versus 2.3 with PyTorch. The company plans to target robotics, edge computing, and cloud deployments, with commercialization expected later this year.