From 2c352c738c7b124f9d1b85df1ac4f56df49f5f53 Mon Sep 17 00:00:00 2001 From: rwkeyes <86691029+rwkeyes@users.noreply.github.com> Date: Fri, 13 Nov 2026 12:19:41 -0500 Subject: [PATCH] docs: the 0.2.13 gain is server-side, not 'entirely in the server' The two trees are not the same binary: the port carries 38 files over the tag, including the engine's TUNE channel. What the battery shows is that the engine measures the same on both sides (within 1.1 % on eight legs), so the claim is now stated that way instead of implying identical binaries. --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 433664cb150..0e1082b1f1d 100644 --- a/README.md +++ b/README.md @@ -215,8 +215,8 @@ Same card, same engine version (0.1.59), same model and pack, same server config **The short version.** On a stream of fresh prompts the first token arrives **11–22 % sooner** — 739 ms against stock's 949 ms at one stream — and at four streams the same 45-second window finishes **three more answers**. Nothing -is traded for it: prefill and decode are within 1.1 % of stock on all eight engine-level legs, because the engine -version is the same on both sides and the difference is entirely in the server. +is traded for it: prefill and decode are within 1.1 % of stock on all eight engine-level legs, and the gain is a +server-side one — the engine measures the same on both sides.

Time to first token, Dirigo 0.2.13 vs stock 0.2.13 on an Intel Arc Pro B70: 739 vs 949 ms at one stream, 2622 vs 3163 at two, 6688 vs 7500 at four, 9641 vs 9659 at eight. Engine-level prefill and decode are within 1.1 percent on all eight legs.