From 8849438bce4f1782fd46281b1d4eacfb18440be3 Mon Sep 17 00:00:00 2001 From: m3ultra Date: Sun, 19 Jul 2026 21:41:28 +1000 Subject: [PATCH] =?UTF-8?q?benchmarks:=20Qwen-Layered=20cross-bench=20?= =?UTF-8?q?=E2=80=94=20m3ultra=20243.0s=20(1.50x=20vs=20m1),=20both=20Ultr?= =?UTF-8?q?as=20serving=20from=20baked=20q8?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-Authored-By: Claude Fable 5 --- BENCHMARKS.md | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/BENCHMARKS.md b/BENCHMARKS.md index e688b12..21c8c02 100644 --- a/BENCHMARKS.md +++ b/BENCHMARKS.md @@ -362,3 +362,14 @@ q4/q6 quality ladder, m3ultra cross-bench. Remaining ladder (untested): q6/q4 quality/speed, 1024-res, fewer-steps floor (10?), m3ultra cross-bench (~1.4× compute but needs 34GB disk — m3 disk currently tight), MODELBEAST operator wrapper (qwen_layered_local) once defaults settle. + +### Cross-bench addendum: m3ultra joins (same baked q8, 20 steps, 640, 4 layers) + +| box | wall | note | +|---|---|---| +| m3ultra | **243.0 s** | 1.50× vs m1 — matches fleet-matrix TFLOPS ratio (1.44×) | +| m1ultra | 365.6 s | | + +Both Ultras now serve Qwen-Image-Layered from the same 34GB baked-q8 +artifact (m3: ~/qwen-layered/, m1: ~/qwen-layered-staging/) + the same +Gitea fork clone. Peak 36.7GB either box. m3 disk after install: ~82GB free.