majentik/MERaLiON-3-10B-TurboQuant-MLX-8bit

Card accuracy pass 3: remove brand-comparison claim tables

Card accuracy pass 2: remove unmeasured speed claims, honest brand labels

Card accuracy sweep: honest brand labeling, remove dead links, upstream KV tip

docs: upstream-first KV-cache guidance (q8_0/q4_0, mainline Hadamard rotation); fork demoted to experimental

docs: Tier 2 polish — variant matrix + quant trade-off

chore(card): enrich YAML frontmatter (pipeline_tag, language, library_name, inference)

Re-upload with properly quantized decoder (encoder/adaptor bf16)

Add MLX quantized model with KV cache compression

initial commit