-
Add validated RTX PRO 6000 (SM120) serving recipe
-
Final card: matched MTP-on head-to-head (125.7 vs 116.3 vs 94.4 tok/s @ c1, best TTFT), FP8-parity evals, clean presentation
-
Correction resolved: fair matched-MTP re-run on 8xH200 (ours 125.7 vs NVFP4+MTP 116.3 vs FP8+MTP 94.4 @ c1) — honest +8%/+33% interactive lead, ~11% behind at c32
-
Correction 2026-07-01: retract non-apples-to-apples MTP-on-vs-off throughput comparisons, keep verified quality and footprint claims
-
Update README.md
-
Upload README.md with huggingface_hub
-
Add files using upload-large-folder tool
-
Upload README.md with huggingface_hub
-
initial commit
