sudoingx/Ternary-Bonsai-2-27B-PTQ1_0-MTP-GGUF

Card: drop the two mentions of the #221 author

v1.1: the RTX 50 bundle (CUDA 12.8, sm_120), SHA256SUMS and the card

Remove the superseded dcc3be7 bundle

Docs: point at the 285542d bundle

Docs: checksums for the 285542d bundle

Add: bundle at 285542d, prefill 267 to 522 tok/s with the 214 tile loader underneath

Remove the superseded 8971d7b bundle

Docs: point at the dcc3be7 bundle

Docs: checksums for the dcc3be7 bundle

Add: bundle rebuilt at dcc3be7, serve scripts pass reasoning-effort medium

Docs: reasoning-effort medium, the xhigh default eats the output budget

Docs: qwen3_5 tag

Fix: the 41.8K row, n-max 1 at every depth

Docs: tarball row, checksums

Add: SHA256SUMS

Add: prebuilt Linux sm86/sm89 llama.cpp bundle

Docs: model card

Add: the lean MTP file

Add: the fat MTP file

initial commit