Splash is our open-source inference engine for Apple silicon, built around the model rather than around a model zoo. On a 48 GB M5 Pro it generates 210 tokens/s on Qwen3.6-35B-A3B, reopens a cached 32K context in 123 ms, and serves four concurrent requests at 357 tokens/s combined.
Most inference engines are built to run any model for maximum flexibility. Our approach turns this assumption upside down: the engine is built around the model for maximum efficiency . Its kernels, draft model, and memory plan are specialized for the model it serves. This design runs our industry-leading inference platform on data-center GPUs, where we are the fastest provider on Artificial Analysis for five frontier open models.
Today we bring the same technology to Apple silicon with Splash , our open-source inference engine for the Mac. It delivers 2× the decode speed of the next-fastest engine we measured on Qwen3.8-27B and stays ahead at every context length we tested, to 32K tokens. At four parallel subagents, the speedup increases to almost 4× .
