philipjohnbasile/Qwen3.6-27B-Fable-Fusion-711-MTPLX-8bit

Clarify release scope and connect the model portfolio

Replace single-sample throughput figures with characterised distributions: ranges, trial counts, methodology, and the 4-bit context-copy bimodality

Withdraw incorrect throughput correction (measured under load); restore original figures, flagged unverified

Correct throughput: republish measured AR/D1/D2/D3 and per-position acceptance

Correct behavioral claims and link validated v2 artifact

Credit nightmedia F32 source; add independent cosine validation

Add LICENSE and NOTICE (Apache-2.0 compliance)

Add LICENSE and NOTICE (Apache-2.0 compliance)

Correct per-variant draft acceptance (measured, identical prompt)

Fable-Fusion-711 MLX conversion from Q8_0 MTP GGUF

initial commit