docs: remove preliminary benchmarks pending official eval

fix: update links for merged branches, add 4bit quant scores (33.6%)

Update README.md

Update README.md

docs: remove MLX from main, compact Use section with links

docs: add title attributes with run history to benchmark links

docs: merge best-per-category scores into grid, link to bf16 methodology

docs: remove best-per-category from main, keep raw grid only

docs: add MMLU-Pro best-per-category table (46.4%), 5-shot results

docs: add 5-shot CoT methodology note, mlx_lm verification

docs: fill Lemer(1,1) scores across all 14 categories

Update README.md

Update README.md

Update README.md

docs: remove MMLU-Pro from training datasets (benchmark only)

docs: add LEM-research dataset reference

docs: link quant table headers to branches

docs: clean header, move methodology below table

docs: compact column headers to (Think,Temp) notation

docs: add 4-dimension math benchmarks, dataset link, remove opinions

docs: add benchmarks grid + rapid-mlx usage

Update README.md

Update README.md

Update README.md

Update README.md

Update README.md

Update README.md

Update README.md

Fix: add quantized base_model relation for model tree

Add YAML metadata to model card

Add model card