From c44892e5fe933d7d7dc41dbfc6110bb3797b9647 Mon Sep 17 00:00:00 2001
From: balancedscorpion <175030879+balancedscorpion@users.noreply.github.com>
Date: Sat, 17 Oct 2026 22:41:09 +0000
Subject: [PATCH 1/2] docs: measure the ModelOpt FP4-Local checkpoint on one
 GB10

The recipe's measurements covered only the MLX affine checkpoint. Add the ModelOpt one, measured with
beec1aad6 built unpatched for sm_121: decode and concurrency from bench_openai and bench_concurrent, cold
first tokens from bench_cache with exact token arrays, the served exactness checks, the seams at fixed
histories, and what the 128 GB machine admits.
---
 docs/recipes/flash-next-cuda.md | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/docs/recipes/flash-next-cuda.md b/docs/recipes/flash-next-cuda.md
index 1746cb63b4..613bf6d3da 100644
--- a/docs/recipes/flash-next-cuda.md
+++ b/docs/recipes/flash-next-cuda.md
@@ -131,6 +131,29 @@ candidates are close: 32 greedy tokens equal it on four of `zig/tests/cuda/flash
 token 15 on the 3,001-token one. Over 4,096 tokens of this repository's documentation, the next-token perplexity is
 6.168 on the captured set and 6.178 on ours.
 
+### The ModelOpt checkpoint on one GB10
+
+`nvidia/Qwen3.8-Flash-Next-FP4-Local` (revision `b9ce18a6`) on a GB10 (128 GB, driver 580.178, CUDA 13.3), greedy,
+`--prompt-cache-gib 0`, the tools above, three runs of the defaults at a 32,768-token window alternated with other
+settings in one session:
+
+| | defaults |
+| --- | --- |
+| One stream, drafted, code / chat | 64.1-64.6 / 40.7-41.1 |
+| drafted, 1 / 4 / 8 streams, greedy code | 64.1-64.4 / 44.9-45.2 / 66.2-66.8 |
+| drafted, 1 / 4 / 8 streams, greedy chat | 40.6-41.1 / 35.0-35.2 / 52.4-52.8 |
+| Cold first token at 2,048 / 8,192 / 32,768 / 65,536 / 131,072 tokens (s) | 1.83 / 7.11 / 29.2 / 60.1 / 131.7 |
+
+Every concurrent reply equalled its solo run and every drafted one its `"draft": false` run. A reply cancelled
+mid-stream, or sent again after a restart, equalled a fresh one, and a second conversation turn resumed from the prompt
+cache (`--no-drafts`) equalled the same turn computed without it. Prompt rows equal one-token rows bit for bit at seven
+fixed histories of 20 to 6,496 tokens. Three streams draft at a time; the rest share plain rounds.
+
+The engine holds 73.2 GiB and keeps a 12.2 GiB reserve. A 32,768-token window then admits 7 or 8 streams of 1.21 GiB
+and a 131,136-token window (the cold prompts above) 2 of 4.03 GiB; a 262,144-token stream takes 7.79 GiB, which fitted
+in one start of four on this machine. After the 131,072-token prompt the process had peaked at 6.5 GB of host and
+83.0 GB of device memory. Absolute speeds on this machine moved by about a tenth between sessions.
+
 ## Not yet
 
 - The dense projections run a copy of the Python engine's group-32 lane matmul (`qmm.cu`); the core's `qmm_group` and

From 1613ff9d708f49768caa87d82915c36865253fb9 Mon Sep 17 00:00:00 2001
From: akol1 <15747200+akol1@users.noreply.github.com>
Date: Mon, 19 Oct 2026 12:23:34 -0700
Subject: [PATCH 2/2] docs: date the GB10 measurements and keep only the
 receipted concurrency row

The #952 write-up records the concurrent aggregate for code at 1 / 4 / 8 streams,
not for chat, so the chat row is dropped rather than carried without a receipt.
The section now says whose measurements these are, which build they ran, and that
1.1.2 has not remeasured them.
---
 docs/recipes/flash-next-cuda.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/docs/recipes/flash-next-cuda.md b/docs/recipes/flash-next-cuda.md
index 613bf6d3da..5e14bc9912 100644
--- a/docs/recipes/flash-next-cuda.md
+++ b/docs/recipes/flash-next-cuda.md
@@ -135,13 +135,13 @@ token 15 on the 3,001-token one. Over 4,096 tokens of this repository's document
 
 `nvidia/Qwen3.8-Flash-Next-FP4-Local` (revision `b9ce18a6`) on a GB10 (128 GB, driver 580.178, CUDA 13.3), greedy,
 `--prompt-cache-gib 0`, the tools above, three runs of the defaults at a 32,768-token window alternated with other
-settings in one session:
+settings in one session. These are @balancedscorpion's measurements on `beec1aad6`, the #952 branch build before
+1.1.0, and have not been remeasured on 1.1.2:
 
 | | defaults |
 | --- | --- |
 | One stream, drafted, code / chat | 64.1-64.6 / 40.7-41.1 |
 | drafted, 1 / 4 / 8 streams, greedy code | 64.1-64.4 / 44.9-45.2 / 66.2-66.8 |
-| drafted, 1 / 4 / 8 streams, greedy chat | 40.6-41.1 / 35.0-35.2 / 52.4-52.8 |
 | Cold first token at 2,048 / 8,192 / 32,768 / 65,536 / 131,072 tokens (s) | 1.83 / 7.11 / 29.2 / 60.1 / 131.7 |
 
 Every concurrent reply equalled its solo run and every drafted one its `"draft": false` run. A reply cancelled