eggdog100/Qwen3.6-35B_Zenith

Update README.md

Add GGUF download table + new-arch runtime caveat

Delete gguf/Zenith-mmproj-f32.gguf with huggingface_hub

Delete gguf/Zenith-mmproj-f16.gguf with huggingface_hub

Delete gguf/Zenith-Q4_K_M.gguf with huggingface_hub

Delete gguf/Zenith-Q6_K.gguf with huggingface_hub

Delete gguf/Zenith-Q8_0.gguf with huggingface_hub

Delete gguf/Zenith-F16.gguf with huggingface_hub

Qwen3.6-35B_Zenith-IQ2_XXS.gguf (no-MTP runtime-fixed)

Qwen3.6-35B_Zenith-Q4_K_M.gguf (no-MTP runtime-fixed)

Qwen3.6-35B_Zenith-Q6_K.gguf (no-MTP runtime-fixed)

Qwen3.6-35B_Zenith-Q8_0.gguf (no-MTP runtime-fixed)

Qwen3.6-35B_Zenith-F16.gguf (no-MTP runtime-fixed)

Qwen3.6-35B_Zenith-mmproj-f32.gguf

Qwen3.6-35B_Zenith-mmproj-f16.gguf

Eval vs independent base (no regression) + style samples; drop unused benchmarks

Add Zenith-Q4_K_M

Add Zenith-Q6_K

Add Zenith-Q8_0

Add Zenith eval scores (4/6 benchmarks) + protocol caveat

Add gguf/Zenith-F16.gguf

Add gguf/Zenith-mmproj-f32.gguf

Add gguf/Zenith-mmproj-f16.gguf

Add LoRA adapter (r=32)

Add Zenith merged bf16 weights (650-step LoRA SFT, vision preserved)

Eval table: official reference + Zenith (drop redundant base column)

Add official base-model reference scores + eval methodology note

Add model-card YAML metadata (base_model, license, tags)

Add model card: capability SFT (LoRA, vision-preserved); eval results TBD

initial commit