Model card: document sampling defaults fix

Add generation_config.json (temperature=1.0, top_p=0.95, ZAI defaults)

Drop internal index.json fix note from model card

Model card: trim the 2026-06-19 fix note to the essentials

Model card: note 2026-06-19 MTP/index fix (re-pull index.json for working speculative decoding)

Fix index.json: map layer-78 (MTP/nextn) to shard model-00041 — restores EAGLE speculative decoding (was orphaned/unindexed -> 0.1% accept)

Update model card: friendly factual lead, H100 KV note, tightened wording

Card: replace em-dash table placeholders with n/a

Card: plain professional prose (drop pull-quote banners), restore published reference benchmarks

Card: prominent 1M-context + no-loss claims; clarify param count (4-bit packing, same ~750B); align with official GLM-5.2 without internal refs

Restructure model card: quantization summary on top, original GLM-5.2 card appended

Add model card

Upload model-00041-of-00041.safetensors with huggingface_hub

Add files using upload-large-folder tool

initial commit