Splash fork optimised for M5 Max: ~1.5× faster (1.25× single request)

A developer has optimized Inco's Splash engine for M5 Max chips, resulting in a new fork called Splish. This optimization, achieved using Opus 5.5, makes Splish approximately 1.5 times faster overall, with single requests showing a 1.25 times improvement. Specifically, Swift-1.5 saw a 20% increase in performance, going from 74.7 to 89.8 tok/s, largely due to enabling a tuner present in Splash's source code but not in the packaged app. The developer is seeking results from other machines.