retitle: seven tiers, 4.55-8.1 bpw

add q5f/q6f tiers: README.md

add q5f/q6f tiers: CHECKSUMS.md5

add q5f/q6f tiers: qwen36-27b-mtp-q6f.q27

add q5f/q6f tiers: qwen36-27b-mtp-q5f.q27

CHECKSUMS: add q8-v1

q8-v1 tier: full Q8_G128 (28.45 GB, ~8.1 bpw) -- 48GB+ cards; UNVALIDATED pending the ladder on 48GB-class hardware

Upload qwen36-27b-mtp-q4s.q27 with huggingface_hub

Upload README.md with huggingface_hub

Upload CHECKSUMS.md5 with huggingface_hub

model card: q6k tier section

CHECKSUMS: add q6k

q6k-v1 tier: 6.8 bits/param (q6 + ffn_gate Q8), PPL 7.9127

model card: q6 tier section + revise stale Qwopus +35% claim to +6% real-traffic

CHECKSUMS: add q6

q6-v1 tier: 6.0 bits/param (v1.4 + ffn_down Q8), PPL 7.946 vs Q5_K_M bar 7.918 (+0.35%; 5.25bpw is +1.55%)

Upload qwen36-27b-mtp.q27 with huggingface_hub

Upload CHECKSUMS.md5 with huggingface_hub

Upload qwen36-27b-mtp.tok with huggingface_hub

Upload README.md with huggingface_hub

initial commit