Qwen3.6 mini GGUFs trim MTP grafting

Useful niche plumbing: it does not make inference easier by itself, but it removes the most annoying part of the MTP grafting workflow for people already managing local model libraries.

• –The payoff is bandwidth and setup time, not new capability; these are compatibility shims for an existing conversion script.
• –The author claims SHA256 parity against outputs made from the full models, which is the right validation for this kind of utility.
• –Scope is narrow: only the two tested Qwen3.6 variants are covered, and the post itself warns that other model variants may fail.
• –The artifact depends on an unstable MTP ecosystem, so treat it as a convenience layer rather than something to archive as canonical.