The post benchmarks newly released Apple Silicon inference engines Splash 1.4.0 and Kapitsa 0.2 (featuring the KFlash speculative decoding method) against other engines such as MLX on Qwen 3.8 27B/4B/35B models. Testing was performed on actual MacBook Pros (M4 Pro 48 GB 2024 and M1 Pro 16 GB 2021), showing token rates around 50 tok/s for Splash (DFlash2) and approximately 65 tok/s for Kapitsa (KFlash). These results provide a comparative performance view for local inference on Apple Silicon hardware.