Launches
Humanizer's DSpark draft reports 1.54x rewriting throughput on Apple silicon
Community reports 5 sources
H
Humanizer Draftfit DSpark
Model Oct 15, 2026
A target-specific speculative draft for accelerating the Humanizer rewriting model through a tested Apple silicon MLX research runtime.
Visit primary link
2 sources attached
What happened lately
Humanizer Draftfit DSpark timeline
Humanizer Draftfit DSpark proposes tokens that the Humanizer target model verifies. It is a speculative draft tied to that target, rather than a standalone rewriting model. The author’s tested path uses Apple silicon and a patched MLX runtime; vLLM and SGLang serving remain unverified for this checkpoint.
The model card publishes a short-output pilot, checkpoint comparisons and a runner with a target-only token check. Its measurements describe the author’s specific runtime and input set, not a general speed guarantee.