Youssofal/Ternary-Bonsai-2-27B-MTPLX-Optimized-Speed

Full model name in the tagline, Prism ML's MLX figure next to MTPLX's, and list the pack as a merge of its Prism ML and Qwen parents

MTPLX 2.12.0 is out: the card no longer calls it the next release

Model card in the MTPLX card format: what MTPLX runs, measured speed, how to use it

Record the exactness measurement against Prism ML's runtime: mean KL 4.0e-6 in float16, 1.1e-7 with float32 auxiliary tensors

Plain card for the 2.12.0 release: medium reasoning by default, measured speed and a short memory table

Publish Bonsai 2 with MTP support for the upcoming MTPLX 2.12.0 release

initial commit