Clarify RTX PRO 6000 supports both TP=2 (2 GPUs) and TP=4 (4 GPUs) deployments

Unify model card to canada-quant family template (TL;DR, family table, honest limitations, changes)

Consolidate B300 + RTX 6000 Pro into headline tables (no footnote)

Add RTX PRO 6000 Blackwell (SM 12.0) serving section with measured numbers

README: B300 HBM3e 288 GB spec

Correct BF16 source size (V3 carryover 1.3 TB → V4 ~600 GB)

README: license=mit, honest calibration disclosure, AIME wall-clock vs decode disambiguated, absolute doc links

License: correct frontmatter to MIT (matches upstream DeepSeek-V4-Flash)

README: scrub DSV4-Pro mentions; replication template now architecture-agnostic

README: humble rewrite (no emojis, no superlative framing)

Add files using upload-large-folder tool