ravenx-mtp-drafter. Multi-Token Prediction drafter for speculative decoding — model-agnostic, MLX-native. Extracted architecture from Google's hidden Gemma 4 MTP heads. Train custom drafters for ANY model on Apple Silicon.

github.com/DeadByDawn101/ravenx-mtp-drafter

Vaya's read on this project

Problem, audience, market, and the verdict — sign in to see it.

Updates

No recent activity.