From a1c4701d7ec7ef52a4087f63e3cb6f177c160a71 Mon Sep 17 00:00:00 2001 From: Frank Date: Sun, 4 Oct 2026 18:17:43 -0700 Subject: [PATCH] =?UTF-8?q?docs(frank-small):=20Q8=5F0=20rung=20evaluated?= =?UTF-8?q?=20and=20closed=20=E2=80=94=20floor=20break=20plus=20-24%=20dec?= =?UTF-8?q?ode?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../LLM/frank-small-llama-mimo-q6k-2x200704.md | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/serverAppConfigs-docs/LLM/frank-small-llama-mimo-q6k-2x200704.md b/serverAppConfigs-docs/LLM/frank-small-llama-mimo-q6k-2x200704.md index 96c302a..44904a8 100644 --- a/serverAppConfigs-docs/LLM/frank-small-llama-mimo-q6k-2x200704.md +++ b/serverAppConfigs-docs/LLM/frank-small-llama-mimo-q6k-2x200704.md @@ -54,6 +54,16 @@ field, never `total − used`: the driver holds ~450 MiB of reserve the subtraction hides (that error once overstated this card's slack by a half-GiB). +**Quant ceiling, evaluated and closed 2026-10-03 (Frank: stay here).** +`Q8_0` is the only rung above this one on the Hub (9,545,979,424 B = 9,104 +MiB, +1,671 over Q6_K): at the live pool it would leave 1,178 MiB free — +under the floor — and the calibrated bandwidth law (it predicted this file's +94.4 t/s from Q5_K_M's 104.84 within 1 %) puts decode at ~72 t/s, a further +−24 % for the near-lossless rung. Fitting it was priced two ways (trim the +pool to 2 × 182,272 — which also breaks frank-large's 183,501-token +compaction-prompt fit by 1,236 tokens — or revert V to `q8_0`), and both +were declined. Re-open only on evidence the Q6_K residual matters. + **Pool prices on this card come from boots, never formulas** — the measured boot-pair table lives in [[frank-small-llama-mimo-q5km]] (*Shape and KV contract*) and still prices every rollback shape.