frank-secondary: swap Gemma4-26B-A4B Q4_K_M -> Qwen3.5-9B-Uncensored Q6_K 2x200,704

Capacity trade: the 9B dense-hybrid (qwen35: 32 layers = 8 full-attention + 24 Gated DeltaNet, 4 KV heads x 256 dim, native 262,144, MTP-less, natively multimodal) costs ~7 GiB of weights where the 26B MoE cost ~16.6 GiB. The freed room buys back the second conversation slot and exact KV: --ctx-size 401408 --parallel 2 (two 200,704 windows), f16 K + f16 V, MTP off (the file carries no nextn/mtp tensors - flag-only enable is impossible), vision via the publisher's BF16 mmproj, ubatch 1024 (the CUDA decode-optimal value), verbosity 4, publisher sampler recipe (0.6/0.95/20/0) in models.yml. Hosts with the MiMo-V2.6-Distill Q6_K boots at this geometry: 21,276 MiB used / 2,849 free at ubatch 1024, 40,960 tokens of fp16 KV at ubatch 2048. Compaction stays routed here; window math is unchanged (maxTokens 16,384 keeps the worst-case harness request inside the slot). Gemma4 payload (Q4_K_M + mmproj + MTP assistant) stays on disk as the rollback; its runbook is retired in place, keeping the attention_k_eq_v pair-billing law. Co-Authored-By: omp <omp@franksdank.cloud>
2026-10-07 14:47:32 -07:00
1 parent b51e078e36
commit 1d062f06a3
@@ -27,7 +27,7 @@ merge absorbed the former `vllmScripts` repo, history preserved).
 |--------|---------|---------|-------|------|
 | Dankserver | `frank-primary` | **radiance** (StillDeadcode v1.0.7 line; image `stilldeadcode/radiance@sha256:4c3f484564…` digest-pinned INDEX digest, 287 MiB, `libr4d` gfx1201 kernels, plain `podman run`, root systemd unit) | `Qwen/Qwen3.8-Flash-Next-FP8` official (172.8 GB, 131 shards, sha256-verified) converted **no-recipe** to `qwen4exp-fp8-native.rad` (177,109 MiB measured = 173.0 GiB) — **native modelopt block-FP8, no requantisation** — on **BOTH R9700s TP=2** (`--tp-wire wht6`), experts in a **40 GiB/rank pinned host pool** + VRAM cache (`--placement expert_tiered`, `--expert-vs-cache-ratio 0.82` = the 99 %-hit point), **PLE 47.7 GiB pinned to host RAM** (`--ngram-placement ram`; root unit + `LimitMEMLOCK=infinity`), **fp8 KV 7.32 KiB/token**, **MTP spec decode depth 3**, *…
 | Dankclaw | `frank-large` | **halogen-flash-server 0.16.2** (`ghcr.io/peonist-ai/halogen-flash-server@sha256:0c61bf84…`, digest-pinned, plain `podman run` via `scripts/serve-frank-large.sh`, root systemd unit) | Quat3rnion/halogen-swift1.5-qwen3.8-flash-next-v2-abliterated **`qwen38-flash-next-v2-swift15-abliterated.hgn`** (66,687,678,464 B = compacted r2 of the card's 89,775,868,032 B original, ~62 GiB resident; the v2 `.hgn` format loads in NO other engine), **+ REQUIRED `qwen38-flash-next-ngram.hgn`** (51,200,246,144 B, peonist @5cc17cea — read beside the checkpoint) **+ vision tower** (897,916,416 B, @e053f488 — **vision ON**, `image_url` parts only, ≥256 px probes), **MTP spec decode ALWAYS-ON (`HALOGEN_SPEC_ADAPT=0` — the adaptive default disables the head per request and HALVES decode)**, MTP head INSIDE the checkpoint (`HALOGEN_MTP_HEAD` unset; no sidecar fetched), **2 slots × 200,192 ctx** (KV pool 400,384 positions ≈ 10.7 GiB at ~28 KiB/position; the 27B slots>1-disables-drafting law does NOT carry over — batching and speculation coexist, the card's quickstart ships 2 with the head), **server-side sampler 1.0/0.95/20** (card recipe; models.yml extraBody matches), coupled cap/timeout 65536/7200, prompt cache ON by default (warm 100k turn ≈ 2 s vs ≈ 88 s cold) | 7770 |
-| Dankclaw | `frank-secondary` | llama.cpp master (0.5.0-dev, commit `5fc4f3c`), CUDA sm_86 build (root systemd unit, relinked) | HauhauCS/Gemma4-26B-A4B-QAT-Uncensored **Q4_K_M** (25.2 B-param MoE, ~4 B active, 16,780,192,888 B reported by the engine; 30 layers = 25 sliding-window (`n_head_kv` 8, window 1024) + 5 full-attention (indices 5/11/17/23/29, `n_head_kv` 2)) + the publisher's own **BF16 mmproj** for **vision**, on the RTX 3090, **ONE 200,704-token slot** (`--parallel 1`; two 200,704 slots do not fit 24 GiB), **f16 K + q8_0 V** under flash attention, batch 4096 / **ubatch 1024**, `--log-verbosity 4`, **MTP OFF** (the assistant is a SEPARATE file and its draft KV is built against the target's full window — 23,806 / 319 MiB free with it loaded). Measured boots: **q8_0 K/ubatch 2048 = 21,890 used / 2,235 free; f16 K/ubatch 1024 = 21,868 / 2,257** (the +2,161 MiB of f16-K KV is paid by the ubatch cut). `attention_k_eq_v` is a compute-graph mapping only — the cache bills the K+V pair **twice** (6.125 B/elem at f16 K + q8_0 V; verified exactly against the engine's own `llama_kv_cache` lines). Cutover 2026-10-07 from Swift-1.5 RCO-IQ3_XXS 2×114,688. All three endpoints route harness **compaction** here. | 7771 |
+| Dankclaw | `frank-secondary` | llama.cpp master (0.5.0-dev, commit `5fc4f3c`), CUDA sm_86 build (root systemd unit, relinked) | HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive **Q6_K** (7,359,259,008 B; dense `qwen35` hybrid: 32 layers, 8 full-attention + 24 Gated DeltaNet, 4 KV heads × 256 dim, vocab 248,320, native 262,144, **no MTP tensors in the file**) + the publisher's own **BF16 mmproj** (921,704,448 B) for **vision**, on the RTX 3090, **TWO 200,704-token slots** (`--ctx-size 401408 --parallel 2`), **exact f16 K + f16 V** under flash attention (32 KiB/token over the 8 full layers = 12.25 GiB pool), batch 4096 / **ubatch 2048** (measured CUDA optimum), `--log-verbosity 4`, **MTP OFF** (forced: no `nextn` tensors exist). Priced from the MiMo Q6_K boots (same geometry, same card, same shape): 21,276/2,849 at ubatch 1024, 21,846/2,279 at 2048. Cutover 2026-10-07 (late) from Gemma4-26B-A4B Q4_K_M 1×200,704. All three endpoints route harness **compaction** here. | 7771 |
 Retired endpoints (`frank-npu`, and every prior runtime) are in
 the [runbook index](#runbook-index). The full swap history lives in
@@ -308,12 +308,13 @@ image]`) — `frank-large` returned to vision with the flash-server v2 swap
 27-layer Qwen-VL tower — patching, pad-token expansion and 3-part rotary
 are the engine core; `--mm-max-patches` available),
 and `frank-secondary` through
-the publisher's own BF16 mmproj passed as `--mmproj` — the Gemma4
+the publisher's own BF16 mmproj passed as `--mmproj` — the model is natively
-GGUF itself carries no vision tensors. `frank-secondary`'s `contextWindow` is
+multimodal but the Q6_K GGUF carries no vision tensors (same arrangement the
-**200704** (the one-slot window since the 2026-10-07 Gemma4 swap; `--parallel 1`,
+Gemma4 era used). `frank-secondary`'s `contextWindow` is
-so the pool IS the conversation), the
+**200704** — the PER-SLOT window (`--ctx-size 401408 --parallel 2` since the
-per-slot window, not the model's native 262144: the harness must compact
+2026-10-07 Qwen3.5-9B swap), not the pool and not the model's native 262144:
-before llama.cpp answers a longer prompt with a 400. Note also
+the harness must compact before llama.cpp answers a longer prompt with a 400.
 Note also
 that
 the reasoning channel differs by runtime: llama.cpp AND SGLang (qwen3
 parser, live-verified) emit `reasoning_content`, vLLM emits `reasoning` —
@@ -322,7 +323,7 @@ There is **no audio**
 input anywhere in the fleet: that capability left with the FLM NPU
 endpoint on 2026-09-11, returned 2026-09-17 with the Gemma4 projector's
 audio encoder, and left again with the Gemma4 retirement on 2026-10-01
-(the MiMo tower is vision-only).
+(neither the MiMo nor the Qwen3.5 tower carries audio).
 ### frank-large — halogen-flash-server 0.16.2, Qwen3.8-Flash-Next v2 Swift1.5 ABLITERATED (Dankclaw, ACTIVE)
@@ -419,67 +420,65 @@ that (`scripts/build-rocmfpx.sh`, pin `36e9acd`) is untouched.
 `GGML_CUDA_FA_ALL_QUANTS=ON`) as a **root systemd unit**
 (`scripts/frank-secondary.service`, `relink_unit`ed into
 `/etc/systemd/system/`), serving
-`HauhauCS/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-MTP` at
+`HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive` at **Q6_K**
-**Q4_K_M** (16,796,015,520 B on disk; the engine reports
+(7,359,259,008 B, sha256 `c0ba7beb…` computed locally — the Hub carries no
-16,780,192,888 B, 25,233,142,046 params, vocab 262,144) **plus the
+LFS oid) **plus the publisher's own BF16 mmproj** (921,704,448 B), both
-publisher's own BF16 mmproj** (1,194,827,776 B), both
+`/home/dank/models/Qwen35-9B-HauhauCS-Q6K/`, on port 7771. Cutover
-`/home/dank/models/Gemma4-26B-A4B/`, on port 7771. Cutover 2026-10-07
+2026-10-07 (late) from the Gemma4-26B-A4B Q4_K_M 1×200,704 shape (same day).
 from the Swift-1.5 RCO-IQ3_XXS 2×114,688 shape. The MTP assistant
 (`mtp-gemma-4-26B-A4B-it.gguf`, 251,937,728 B) is on disk, **not loaded**.
-**Architecture: gemma4 MoE hybrid** — 30 layers, 128 experts (8 active),
+**Architecture: qwen35 dense hybrid** — read from the GGUF's own header, not
-n_embd 2816, native 262,144 window. Attention is hybrid per the file's
+the model card: 32 layers, `full_attention_interval = 4` ⇒ **8 full-attention
-`sliding_window_pattern`: **25 sliding-window layers (window 1024,
+layers (3, 7, 11, …, 31) + 24 Gated DeltaNet linear layers**, 16 attention
-`n_head_kv = 8`) + 5 full-attention layers (indices 5, 11, 17, 23, 29,
+heads / **4 KV heads** × 256 dim, `ssm` conv_kernel 4 / state_size 128 /
-`n_head_kv = 2`)**, head_dim 256. Two KV caches: the full-context cache
+group_count 16 / time_step_rank 32 / inner_size 4096, native 262,144, vocab
-scales with the 5 layers, the SWA cache is a fixed 1024-token ring per
+248,320. Tensor types are Q6_K × 250 + F32 × 177 — standard quants only, so
-sliding layer. `attention_k_eq_v: True` is a **compute-graph** mapping
+the stock master build loads it (arch `qwen35` is registered in this build's
-(`gemma4.cpp` maps the absent `wv` to `Vcur = Kcur`) — the cache reserves K
+`src/llama-arch.cpp`). **KV accounting:** the pool scales with the 8 full
-and V and, with `k_eq_v`, reserves the pair TWICE: 6.125 B/elem at f16 K +
+layers only — 8 × 4 × 256 = 8,192 elements/token per cache, so f16 K + f16 V
-q8_0 V, 2.125 B/elem at q8_0/q8_0. The engine's own `llama_kv_cache` buffer
+bills **32 KiB/token** = 12.25 GiB at the 401,408 pool; the 24 DeltaNet
-sizes reproduce that rate exactly, so bill the pair, never K alone.
+layers bill a fixed recurrent state per SLOT, not per token, which is exactly
 why two 200k slots fit where the 26B MoE fit one.
-**The shape (2026-10-07):** `--ctx-size 200704 --parallel 1` = ONE
+**The shape (2026-10-07):** `--ctx-size 401408 --parallel 2` = TWO
-conversation (Frank's 200k cap; the harness `contextWindow` quotes
+conversations of 200,704 (Frank's cap; the harness `contextWindow` quotes
-200704), `--flash-attn on`, **f16 K + q8_0 V** (exact K, FP8-class V),
+the per-slot number), `--flash-attn on`, **f16 K + f16 V** (exact KV),
-batch 4096 / **ubatch 1024** (cut from 2048 to fund the f16-K raise),
+batch 4096 / **ubatch 2048** (the measured CUDA optimum: +4.6 % prefill,
-`--log-verbosity 4` (debug — the KV/buffer forensics land in the journal),
+decode flat, 616 MiB — MiMo-era A/B on this card), `--log-verbosity 4`,
 `--mmproj` for vision, **speculative decoding OFF** (`SPEC_TYPE=none`).
-**Why MTP is off**: Gemma4's MTP is a SEPARATE assistant file (arch
+**Why MTP is off**: it is not a choice — **this Q6_K file carries no
-`gemma4-assistant`), not an in-file head, and the draft context is built
+`nextn`/`mtp` tensors** (verified from the tensor-name list), and MTP is a
-against the target's FULL window — measured cost **~1.9 GiB** (23,806 used
+property of the file, not of the flags. The publisher card advertises "MTP
-/ 319 free with the assistant loaded, vs 21,890 / 2,235 without). The
+support" for the base architecture; the quantized export dropped the head.
 assistant also warns `n_ctx_seq (200704) > n_ctx_train (131072)` — the
 draft runs 53 % past its trained window. The q8_0-K boot leaves 2,235 MiB,
 which the f16-K raise (+2,161 MiB of KV) spends together with the ubatch
 cut; MTP and exact-K cannot coexist on this card.
-**Measured boots** (nvidia-smi, the referee): q8_0 K + q8_0 V, ubatch 2048,
+**Measured boots** (nvidia-smi, the referee): the pricing boots are the
-MTP off: **21,890 used / 2,235 free**; same shape with the MTP assistant:
+MiMo-V2.6-Distill-Qwen-9B pair at this exact geometry — 2 × 200,704 f16 K +
-**23,806 / 319**; **f16 K + q8_0 V, ubatch 1024, MTP off (deployed):
+f16 V: **21,276 used / 2,849 free** (ubatch 1024) and **21,846 / 2,279**
-21,868 / 2,257**. The verbosity-4 ledger for the deployed boot: weights
+(ubatch 2048). This file is 415 MiB lighter in weights ⇒ projected ~21,430 /
-16,002.91 MiB CUDA0 + 577.50 MiB CPU-mapped, KV 3,001.25 (5 full layers) +
+~2,700. The deployed boot ledger: see the runbook. Decode is ubatch-invariant
-306.25 (25 SWA rings) MiB, compute buffers 926.39 + 150.63 MiB. Decode is
+(Swift-era 3090 law). The Q6_K rung costs ~10 % decode vs Q5_K_M (weight
-expected ubatch-invariant (Swift-era 3090 law; unmeasured on Gemma4).
+bandwidth; MiMo-measured 94.4 vs 104.8 t/s).
-**Vision**: the publisher's own BF16 mmproj via `--mmproj`. Probe traps
+**Vision**: the publisher's own BF16 mmproj via `--mmproj` (the model is
-carry over from the Swift era: ≥256 px probe images or the model answers
+natively multimodal; the Q6_K file carries no vision tensors). Probe traps
-"video", and only the OpenAI `image_url` content part reaches the model.
+carry over: ≥256 px probe images or the model answers "video", and only the
 OpenAI `image_url` content part reaches the model.
 **Compaction hub**: `models.yml` routes `compactionModel` for
 `frank-primary`, `frank-large` and `frank-secondary` here — 200,704 is the
 only window in the fleet that fits a worst-case compaction input (~133k
-tokens). Consequence of `--parallel 1`: compaction requests serialize
+tokens). With `--parallel 2` a compaction request no longer serializes behind
-behind live generation on this slot.
+a live generation (the Gemma4 era's `--parallel 1` law).
 Runbook:
-`serverAppConfigs-docs/LLM/frank-secondary-llama-gemma4-q4km-200704.md`.
+`serverAppConfigs-docs/LLM/frank-secondary-llama-qwen35-9b-q6k-2x200704.md`.
-Rollback: revert the swap commit — the Swift payload
+Rollback: revert the swap commit — the Gemma4 payload
-(`/home/dank/models/Swift15-UncIQ3XXS/`), the MiMo Q6_K payload
+(`/home/dank/models/Gemma4-26B-A4B/`), the MiMo Q6_K payload
-(`/home/dank/models/MiMo-Q6K/`) and the Ornith Q8_0 payload
+(`/home/dank/models/MiMo-Q6K/`), the Swift payload
 (`/home/dank/models/Swift15-UncIQ3XXS/`) and the Ornith Q8_0 payload
 (`/home/dank/models/Ornith-Q8/`) all remain on disk.
 ### frank-primary — Radiance (StillDeadcode), Qwen3.8-Flash-Next FP8 native, TP=2 (Dankserver, ACTIVE)
 The **radiance engine** (StillDeadcode, Apache-2.0; image
@@ -1001,7 +1000,7 @@ Active:
 | Service | Runbook |
 |---------|---------|
 | frank-large (flash-server 0.16.2, abliterated Swift1.5 v2 .hgn + MTP always-on, 2×200,192, vision ON, Dankclaw) | `serverAppConfigs-docs/LLM/frank-large-flash-server-v2-swift15.md` |
-| frank-secondary (llama.cpp CUDA, Gemma4-26B-A4B QAT Q4_K_M, 1×200,704 f16-K/q8_0-V MTP-off, Dankclaw) | `serverAppConfigs-docs/LLM/frank-secondary-llama-gemma4-q4km-200704.md` |
+| frank-secondary (llama.cpp CUDA, Qwen3.5-9B-Uncensored Q6_K, 2×200,704 exact f16 KV MTP-off, vision ON, Dankclaw) | `serverAppConfigs-docs/LLM/frank-secondary-llama-qwen35-9b-q6k-2x200704.md` |
 | frank-primary (radiance StillDeadcode, Qwen3.8-Flash-Next-FP8 native, 2×R9700 TP=2, Dankserver) | `serverAppConfigs-docs/LLM/frank-primary-radiance-qwen38-native-fp8.md` |
 | Dankclaw kernel posture (`amd_iommu=off`) | `serverAppConfigs-docs/LLM/dankclaw-iommu-translation-off.md` |
 | Deployment flow + incidents | `serverAppConfigs-docs/LLM/deployment.md` |
@@ -1010,7 +1009,7 @@ Retired (rollback targets and history):
 | Runtime | Host | Window | Runbook |
 |---------|------|--------|---------|
-| halogen 0.1.4 Qwen3.8-27B p1w4d-d2 + DFlash2 (`frank-large`) | Dankclaw | 2026-10-05 → 10-06 (rollback target — payload + fetch script + image digest on disk; owns the prompt-cache pin and FA-tile lessons) | `frank-large-halogen27b-dflash2.md` |
+| llama.cpp CUDA Gemma4-26B-A4B QAT Q4_K_M, 1×200,704 f16-K/q8_0-V MTP-off (`frank-secondary`) | Dankclaw | 2026-10-07 → 10-07 (same-day swap; rollback target — payload on disk; owns the `attention_k_eq_v` pair-billing KV law and the gemma4-assistant drafter-cost boots) | `frank-secondary-llama-gemma4-q4km-200704.md` |
 | ROCmFPX llama.cpp fork Qwen3.8-27B Uncensored Q6_0_ROCMFPX_AGENT + MTP (`frank-large`) | Dankclaw | 2026-10-03 → 10-05 (rollback target, payload + fork build on disk) | `frank-large-rocmfpx-q6agent-uncensored-mtp.md` |
 | tcclaviger vLLM fork Qwen3.8-Flash-Next-Uncensored FP8, TP=2, shape G (`frank-primary`) | Dankserver | 2026-09-29 → 10-05 (rollback target — tag `frank-primary-vllm-shapeG-20261003`; trio + env lore in the tag, image in rootless storage) | `dankserver-qwen38-fp8-expert-offload-tuning.md` |
 | FreeToken PR #587 ROCm, Qwen3.8-Flash-Next-Uncensored-FP8, ONE R9700 (`frank-primary`) | Dankserver | 2026-10-05 → 10-05 (rollback target — tag `frank-primary-freetoken-pr587-20261005`; trio + Dockerfile in the tag, image `localhost/frank-primary-freetoken:pr587-77a7201` and orcarouter payload on disk) | `frank-primary-freetoken-pr587.md` |
@@ -1225,13 +1225,14 @@ deploy_llm_large() {
 #
 # It is a SEPARATE endpoint on port 7771; frank-large keeps 7770. The 3090 is
 # not shared: frank-large's engine runs on the gfx1151 iGPU.
-LLM_SECONDARY_MODEL_FILE="${LLM_SECONDARY_MODEL_FILE:-/home/dank/models/Gemma4-26B-A4B/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf}"
+LLM_SECONDARY_MODEL_FILE="${LLM_SECONDARY_MODEL_FILE:-/home/dank/models/Qwen35-9B-HauhauCS-Q6K/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf}"
 LLM_SECONDARY_RUNTIME="${LLM_SECONDARY_RUNTIME:-/home/dank/GitPackages/llama-master/build-cuda/bin/llama-server}"
-# The projector IS the vision capability here: the GGUF carries no vision
+# The projector IS the vision capability here: the model is natively
-# tensors, and the publisher ships its own BF16 projector in the same repo, so
+# multimodal and the Q6_K file carries no vision tensors — the publisher ships
-# a deploy that starts without this file silently ships a text-only
+# its own BF16 projector in the same repo, so a deploy that starts without
-# endpoint. Checked unconditionally, like frank-large's.
+# this file silently ships a text-only endpoint. Checked unconditionally, like
-LLM_SECONDARY_MMPROJ="${LLM_SECONDARY_MMPROJ:-/home/dank/models/Gemma4-26B-A4B/mmproj-Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-BF16.gguf}"
+# frank-large's.
 LLM_SECONDARY_MMPROJ="${LLM_SECONDARY_MMPROJ:-/home/dank/models/Qwen35-9B-HauhauCS-Q6K/mmproj-Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf}"
 deploy_llm_secondary() {
    # Refuse rather than restart-loop: a unit whose runtime or payload is
@@ -1,27 +1,24 @@
 #!/usr/bin/env bash
 # Download the frank-secondary payload for llama.cpp on the RTX 3090:
-# ajgazin/Swift-1.5-Qwen3.8-27B-Uncensored-Dynamic-MTP-GGUF:RCO-IQ3_XXS
+# HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive at **Q6_K** plus the
-# **with the MTP head in the file** (the -mtp variant), plus the
+# PUBLISHER's OWN vision projector (the BF16 mmproj — the model is natively
-# PUBLISHER's OWN vision projector (mmproj-BF16.gguf, same repo). Public
+# multimodal, so this file is the endpoint's eyes). Public repo — no auth
-# repo — no auth header, and never a token on the command line (it is
+# header, and never a token on the command line (it is world-readable in
-# world-readable in /proc/<pid>/cmdline).
+# /proc/<pid>/cmdline).
 #
 # Fresh downloads, no resume (`curl -C -` once produced a 213 MB oversized
 # file on a 31 GiB payload); -f/--fail-with-body so an HTTP error is not a
-# success, then byte count and sha256. This repo is on Xet storage, so
+# success, then byte count and sha256. The Hub exposes no LFS oid for this
-# the Hub tree/blob APIs expose no LFS oid: the sha256 values below were
+# repo (Xet storage), so the sha256 values below were computed locally after
-# computed locally after the first verified fetch (byte counts match the
+# the first verified fetch — byte counts match the Hub's declared sizes.
-# Hub's declared sizes). Idempotent: complete files are skipped. The
+# A sha of "-" means "first fetch: size-verified, sha printed for pinning".
-# projector is stored under a self-documenting dest name; the URL name
+# Idempotent: complete files are skipped.
 # and sha stay pinned in the table below.
 #
-# 2026-10-06 (Frank, evening): the payload moved to the UNCELERSED
+# 2026-10-07 (Frank): the payload moved to Qwen3.5-9B-Uncensored Q6_K at
-# Swift-1.5-Qwen3.8-27B at RCO-IQ3_XXS (1.56 GiB lighter than the
+# 2 x 200,704 (qwen35 hybrid: 32 layers, 8 full-attention + 24 Gated
-# IQ3_S-mtp it replaced; the lighter quant funds the MTP draft ctx plus
+# DeltaNet, no MTP tensors in this file). Rollback payloads stay on disk:
-# the second 131,072 slot). Rollback payload stays on disk:
+# /home/dank/models/Gemma4-26B-A4B/ (Q4_K_M, the 2026-10-07 era being
-# /home/dank/models/Swift15-IQ3S/ (the IQ3_S-mtp that was live until this
+# replaced), Swift15-UncIQ3XXS/, MiMo-Q6K/, MiMo-Q5KM/, Ornith-Q8/.
 # swap); earlier payloads Ornith-Q8/ and MiMo-Q6K/ as well. Their
 # re-fetch recipes are in git history (predecessors of this script).
 #
 # Usage:
 #   scripts/download-frank-secondary.sh [DEST_DIR] [--verify-only]
@@ -35,15 +32,14 @@ for arg in "$@"; do
        *) DEST="$arg" ;;
    esac
 done
-DEST="${DEST:-/home/dank/models/Swift15-UncIQ3XXS}"
+DEST="${DEST:-/home/dank/models/Qwen35-9B-HauhauCS-Q6K}"
 # repo | hub-name | dest-name | bytes | sha256 (computed locally; Xet repo)
 FILES=(
-  "ajgazin/Swift-1.5-Qwen3.8-27B-Uncensored-Dynamic-MTP-GGUF|RCO/Swift-1.5-Qwen3.8-27B-Uncensored-Dynamic-MTP-RCO-IQ3_XXS.gguf|Swift-1.5-Qwen3.8-27B-Uncensored-Dynamic-MTP-RCO-IQ3_XXS.gguf|10442827392|e13d3e7400b6c7f980ebd30c0edd6be43eb1865b50be8562a9db042706c7d803"
+  "HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive|Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf|Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf|7359259008|c0ba7beb68fd3fe47891bd549486d38dcf62d00817296ea314ad37017f5a4986"
-  "ajgazin/Swift-1.5-Qwen3.8-27B-Uncensored-Dynamic-MTP-GGUF|mmproj-BF16.gguf|mmproj-BF16.gguf|931145888|19acd7fcf4ee504328566eddf0f8e603ef853a205b4830209d2a4908213abdea"
+  "HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive|mmproj-Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf|mmproj-Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf|921704448|05f662501f8bd45607b079723a3e238a4e888fd085a10a53f4057a0e250f6934"
 )
 mkdir -p "$DEST"
 status=0
 for entry in "${FILES[@]}"; do
  IFS='|' read -r repo hubname destname size sha <<<"$entry"
@@ -51,7 +47,9 @@ for entry in "${FILES[@]}"; do
  f="$DEST/$destname"
  have=0
  if [[ -f "$f" ]] && [[ "$(stat -c%s "$f")" == "$size" ]]; then
-    if sha256sum -c --quiet <(printf '%s  %s\n' "$sha" "$f"); then
+    if [[ "$sha" == "-" ]]; then
      have=1
    elif sha256sum -c --quiet <(printf '%s  %s\n' "$sha" "$f"); then
      have=1
    fi
  fi
@@ -73,6 +71,10 @@ for entry in "${FILES[@]}"; do
    status=1
    continue
  fi
  if [[ "$sha" == "-" ]]; then
    echo "VERIFIED  $destname (size only) sha256=$(sha256sum "$f" | cut -d' ' -f1)"
    continue
  fi
  if ! sha256sum -c --quiet <(printf '%s  %s\n' "$sha" "$f"); then
    echo "SHA FAIL  $destname"
    status=1
@@ -1,46 +1,56 @@
 # frank-secondary — llama.cpp (master, CUDA sm_86) serving
-# HauhauCS/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-MTP
+# HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive at Q6_K on the RTX 3090,
-# (Q4_K_M) on the RTX 3090, :7771.
+# :7771.
 #
-# 2026-10-07 (Frank): swap from Swift-1.5-Qwen3.8-27B RCO-IQ3_XXS to
+# 2026-10-07 (Frank): swap from Gemma4-26B-A4B QAT Q4_K_M (one 200,704 slot)
-# Gemma4-26B-A4B Q4_K_M, ONE 200,704 context, f16 K + q8_0 V,
+# to Qwen3.5-9B Q6_K at TWO 200,704-token conversations, EXACT (f16) KV,
-# MTP DISABLED (SPEC_TYPE=none), log verbosity 4.
+# MTP disabled (SPEC_TYPE=none), log verbosity 4.
 #
-# Architecture (read from the file's own header at load): gemma4 MoE — 30
+# Architecture — read from the GGUF's own header (HTTP range read, not the
-# layers, 128 experts (8 active), sliding_window_pattern = 25 sliding
+# model card): arch qwen35, 32 layers, dense 9B, full_attention_interval 4
-# (window 1024, n_head_kv 8) + 5 full-attention (layers 5/11/17/23/29,
+# ⇒ 8 full-attention layers (3,7,11,...,31) + 24 Gated DeltaNet linear layers,
-# n_head_kv 2), head_dim 256, attention_k_eq_v, native 262,144 window.
+# 16 attention heads / 4 KV heads × 256 dim, ssm conv_kernel 4 / state_size
 # 128 / group_count 16 / time_step_rank 32 / inner_size 4096, native 262,144,
 # vocab 248,320, file_type 18 (Q6_K). Tensor-type histogram: Q6_K×250 +
 # F32×177 — standard types only, so the stock master build loads it (no
 # custom-quant lock-in like the ROCmFPX era). NO nextn/mtp tensors exist, so
 # in-file MTP is impossible here; SPEC_TYPE=none is forced by the payload.
 #
-# KV rate law (measured, exact): the cache bills the K+V pair TWICE under
+# KV rate law: 8 full layers × 4 KV heads × 256 dim = 8,192 elements per
-# attention_k_eq_v — 6.125 B/elem at f16 K + q8_0 V, 2.125 B/elem at
+# token per cache. f16 K + f16 V = 32,768 B/token = 32 KiB/token, so the
-# q8_0/q8_0. Full cache = 5 x 2 x 256 x rate x ctx; SWA cache = 25 x 1024 x
+# 401,408-token pool is 12.25 GiB. The 24 DeltaNet layers cost a fixed
-# 8 x 256 x rate. There is no true fp8 KV dtype in this build; q8_0 is the
+# recurrent state per SLOT (not per token) — that is why two 200k slots fit
-# FP8 class.
+# where the 26B MoE fit only one.
 #
-# MTP is OFF: SPEC_TYPE=none strips --spec-type AND --spec-draft-model
+# Feasibility comes from the predecessor's MEASURED boots, not a formula:
-# (spec flags move as a set). The assistant file
+# MiMo-V2.6-Distill-Qwen-9B (identical qwen35 geometry, 7,793,710,624 B) at
-# (mtp-gemma-4-26B-A4B-it.gguf, 0.252 GiB) stays on disk as the drafter
+# 2×200,704 f16 K + f16 V measured 21,276 used / 2,849 free at ubatch 1024
-# payload if MTP is ever re-enabled. Measured VRAM: MTP ON booted
+# and 21,846 / 2,279 at ubatch 2048. This file is 415 MiB lighter in weights,
-# 23,806 used / 319 free; MTP OFF boots 21,890 used / 2,235 free.
+# its BF16 mmproj is 3 MiB heavier ⇒ projected ~21,430 used / ~2,700 free.
 # The verbosity-4 boot ledger is the referee.
 #
 # Ubatch 2048 is the measured CUDA optimum (measured on the MiMo predecessor
 # at this exact arch + shape: +4.6 % prefill at 8.9k depth, decode flat, cost
 # 616 MiB). 1024 is the fallback rung if the boot lands under the 2 GiB floor.
 LLAMA_BIN=/home/dank/GitPackages/llama-master/build-cuda/bin/llama-server
-MODEL_FILE=/home/dank/models/Gemma4-26B-A4B/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf
+MODEL_FILE=/home/dank/models/Qwen35-9B-HauhauCS-Q6K/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf
-MMPROJ_FILE=/home/dank/models/Gemma4-26B-A4B/mmproj-Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-BF16.gguf
+MMPROJ_FILE=/home/dank/models/Qwen35-9B-HauhauCS-Q6K/mmproj-Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf
 ALIAS=frank-secondary
 BIND_ADDR=0.0.0.0
 LISTEN_PORT=7771
-CTX_SIZE=200704
+# CTX_SIZE is the POOL llama.cpp divides across slots: 401,408 / 2 = 200,704
-PARALLEL=1
+# per conversation. MUST stay divisible by PARALLEL (the seam enforces it) and
 # the harness contextWindow quotes the PER-SLOT number — a longer prompt is
 # answered with a 400, never truncated.
 CTX_SIZE=401408
 PARALLEL=2
 BATCH_SIZE=4096
-UBATCH_SIZE=1024
+UBATCH_SIZE=2048
 CACHE_TYPE_K=f16
-CACHE_TYPE_V=q8_0
+CACHE_TYPE_V=f16
 SPEC_TYPE=none
 NGL=99
-
+# Log verbosity threshold (llama.cpp -lv). 4 = debug: KV layout, buffer sizes
-# 2026-10-07 (Frank): verbosity 4 (debug — KV layout, buffers, graph
+# and graph capture land in the journal, which is what the VRAM-fit forensics
-# capture), ubatch 1024, K raised to f16 (V stays q8_0). MEASURED BOOT of
+# read. Levels above 4 flood the journal.
 # this exact shape: 21,868 MiB used / 2,257 MiB free — KV 3,001.25 (full) +
 # 306.25 (SWA) MiB (+2,161 over q8_0 K), compute buffers 926.39 + 150.63
 # (the 2048 rung cost ~616 MiB more, which is what funded the raise). The
 # q8_0-K/ubatch-2048 predecessor booted 21,890 / 2,235.
 VERBOSITY=4
@@ -1,27 +1,27 @@
-#!/usr/bin/env bash
+# Serve HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive (Q6_K) as
 # Serve HauhauCS/Gemma4-26B-A4B-QAT-Uncensored (QAT **Q4_K_M**) as
 # frank-secondary on Dankclaw's RTX 3090 via llama.cpp (master build, CUDA sm_86).
 #
-# 2026-10-07 (Frank): Gemma4-26B-A4B swap with MTP OFF and ONE 200,704-token
+# 2026-10-07 (Frank): Qwen3.5-9B swap at TWO 200,704-token conversations
-# conversation (--parallel 1) — the 24 GiB card cannot fund a second slot at
+# (--ctx-size 401408 --parallel 2) with EXACT f16 KV — the 9B rung frees the
-# this weight, and the MTP assistant's KV (built against the target's full
+# room the 26B MoE Gemma4 shape lacked, so quantized V retires. MTP stays off:
-# window, ~1.9 GiB measured) plus the f16-K raise do not coexist on it.
+# this GGUF carries no MTP tensors at all (verified from its tensor-name
-# Verbosity 4 ships so the VRAM-fit forensics land in the journal.
+# list), so there is no in-file head to draft with. Verbosity 4 ships so the
 # VRAM-fit forensics (KV layout, buffer sizes, graph capture) land in the journal.
 #
-# Architecture: gemma4_26b_a4b — 30-layer MoE (active ~4B), hybrid attention:
+# Architecture: qwen35 — dense 9B, 32 layers with full_attention_interval 4
-# 25 sliding-window layers (fixed 1024-token ring each) + 5 full-attention
+# (8 full-attention layers at 3,7,...,31 + 24 Gated DeltaNet linear layers),
-# layers (5, 11, 17, 23, 29), 8 KV heads x 256 dim, n_ctx_train 262,144. KV
+# 16 heads / 4 KV heads × 256 dim, native 262,144, vocab 248,320. The KV pool
-# therefore scales ONLY with the 5 full layers: ~30.7 KiB/token at f16 K +
+# scales with the 8 full layers: 32 KiB/token at f16 K + f16 V (12.25 GiB at
-# q8_0 V (~6.2 GiB at 200,704). attention_k_eq_v is declared by the file and
+# the 401,408 pool); the DeltaNet layers cost a fixed recurrent state per
-# honoured by the build (bill K-side accounting; the measured boot is the
+# slot. Feasibility is priced from the MiMo-V2.6-Distill-Qwen-9B boots (same
-# referee: 21,890 used / 2,235 free at q8_0/q8_0, MTP off).
+# geometry, same card, same shape): 21,276/2,849 at ubatch 1024, 21,846/2,279
-# Vision comes from the projector file (the GGUF carries no vision
+# at ubatch 2048 — the boot ledger, not arithmetic, is the referee.
-# tensors; the publisher ships its own BF16 mmproj in the same repo), so
+# Vision comes from the publisher's own BF16 projector (the model is natively
-# --mmproj is what gives this endpoint eyes; MMPROJ_FILE="" is the
+# multimodal), so --mmproj is what gives this endpoint eyes; MMPROJ_FILE="" is
-# documented text-only opt-out.
+# the documented text-only opt-out.
 #
-# Sampling: llama.cpp takes it per request (the fleet harness sends its
+# Sampling: llama.cpp takes it per request (the fleet harness sends the
-# recipe from models.yml); greedy is the harness default.
+# publisher recipe from models.yml: thinking temp 0.6 / top_p 0.95 / top_k 20).
 #
 # Tunables come from frank-secondary.env unless FRANK_SECONDARY_ENV_FILE overrides.
 set -euo pipefail
@@ -37,32 +37,35 @@ fi
 # --- defaults (used when the env file does not set them) -------------------
 : "${LLAMA_BIN:=/home/dank/GitPackages/llama-master/build-cuda/bin/llama-server}"
-: "${MODEL_FILE:=/home/dank/models/Gemma4-26B-A4B/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf}"
+: "${MODEL_FILE:=/home/dank/models/Qwen35-9B-HauhauCS-Q6K/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf}"
 : "${ALIAS:=frank-secondary}"
 : "${BIND_ADDR:=0.0.0.0}"
 : "${LISTEN_PORT:=7771}"
-# ctx-size is the POOL llama.cpp divides across slots; with --parallel 1 the
+# ctx-size is the POOL llama.cpp divides across slots: 401,408 / 2 = 200,704
-# pool IS the conversation window. 200,704 is inside the model's trained
+# per conversation, inside the model's trained 262,144. The harness
-# 262,144; the harness contextWindow MUST quote this number — a longer prompt
+# contextWindow MUST quote the PER-SLOT number — a longer prompt is answered
-# is answered with a 400, never truncated.
+# with a 400, never truncated.
-: "${CTX_SIZE:=200704}"
+: "${CTX_SIZE:=401408}"
-: "${PARALLEL:=1}"
+: "${PARALLEL:=2}"
 # Compute-buffer geometry. Ubatch is a prefill knob (decode never moves).
-# 2026-10-07 (Frank): 1024 — cut from 2048 to fund the f16-K raise. The
+# 2048 is the measured CUDA optimum (MiMo era, same arch/shape: +4.6 % prefill
-# logical batch is VRAM- and throughput-neutral at a fixed ubatch.
+# at 8.9k depth, decode flat, 616 MiB). The logical batch is VRAM- and
 # throughput-neutral at a fixed ubatch.
 : "${BATCH_SIZE:=4096}"
-: "${UBATCH_SIZE:=1024}"
+: "${UBATCH_SIZE:=2048}"
-# KV at FP8 class = q8_0 on llama.cpp (1.0625 B/elem => 8.5 effective bits;
+# Exact (unquantized) KV: f16 K + f16 V, restoring the MiMo-era contract that
-# there is no true fp8 KV dtype). Explicit either way so the contract is
+# the 26B Gemma4 shape could not afford. 32 KiB/token at this geometry
-# visible in the argv.
+# (8 full-attention layers × 4 KV heads × 256 dim); 12.25 GiB at the 401,408
-: "${CACHE_TYPE_K:=q8_0}"
+# pool. Explicit either way so the contract is visible in the argv, and
-: "${CACHE_TYPE_V:=q8_0}"
+# --flash-attn stays required for any quantized-V fallback (q8_0 = llama.cpp's
-# Speculative decoding is OFF by default since the Gemma4 swap (Frank,
+# FP8 class; no true fp8 KV dtype exists in this build).
-# 2026-10-07): Gemma4's MTP is a SEPARATE assistant file (arch
+: "${CACHE_TYPE_K:=f16}"
-# gemma4-assistant), not an in-file head, and its KV is built against the
+: "${CACHE_TYPE_V:=f16}"
-# target's full window — ~1.9 GiB measured, which the f16-K raise cannot
+# Speculative decoding is OFF: this Q6_K file carries NO nextn/mtp tensors, so
-# afford. SPEC_TYPE=none strips BOTH speculative flags (flags move as a
+# there is no in-file head (the law: MTP is a property of the file, not of the
-# set); naming a drafter requires SPEC_DRAFT_FILE alongside SPEC_TYPE.
+# flags). SPEC_TYPE=none strips BOTH speculative flags (flags move as a set);
 # naming a drafter requires SPEC_DRAFT_FILE alongside SPEC_TYPE — the Gemma4
 # assistant on disk is arch gemma4-assistant and cannot drafter-match qwen35.
 : "${SPEC_TYPE:=none}"
 : "${SPEC_DRAFT_N_MAX:=3}"
 : "${SPEC_DRAFT_FILE:=}"
@@ -73,7 +76,7 @@ fi
 : "${VERBOSITY:=3}"
 # Vision projector. No colon on purpose: an explicitly empty value means
 # "text-only, on purpose" and must NOT be replaced by the default path.
-: "${MMPROJ_FILE=/home/dank/models/Gemma4-26B-A4B/mmproj-Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-BF16.gguf}"
+: "${MMPROJ_FILE=/home/dank/models/Qwen35-9B-HauhauCS-Q6K/mmproj-Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf}"
 # Fail loudly rather than half-loading: a missing file is a download that
 # has not finished, and Restart=always would turn it into a crash loop that
@@ -1,10 +1,11 @@
 # frank-secondary — llama.cpp CUDA, Gemma4-26B-A4B Q4_K_M, 1×200,704 (Dankclaw)
-**ACTIVE since 2026-10-07.** RTX 3090, port 7771, root systemd unit
+**RETIRED 2026-10-07** — same day it went live. Superseded by
-`frank-secondary.service`, served by `scripts/serve-frank-secondary.sh` +
+[[frank-secondary-llama-qwen35-9b-q6k-2x200704]] (Qwen3.5-9B Q6_K at
-`scripts/frank-secondary.env`. Supersedes
+2 × 200,704 with exact KV); the payload stays at
-[[frank-secondary-llama-swift15-unc-iq3xxs-mtp-114688]] (Swift-1.5-Qwen3.8-27B
+`/home/dank/models/Gemma4-26B-A4B/` as the rollback. This page keeps the
-RCO-IQ3_XXS-mtp, 2×114,688, in-file MTP).
+Gemma4 KV-billing law (the `attention_k_eq_v` pair-billing measurement) and
 the drafter-cost boots, both of which outlive the model.
 ## Payload
@@ -0,0 +1,113 @@
 ---
 type: reference
 title: "frank-secondary on Dankclaw — llama.cpp, Qwen3.5-9B Q6_K, 2 x 200704, f16 KV (ACTIVE)"
 description: "The ACTIVE frank-secondary runbook (since 2026-10-07) — llama.cpp master (CUDA sm_86) serving HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive at Q6_K on the RTX 3090: two conversations of 200,704 tokens with exact f16 KV, vision through the publisher's BF16 projector, MTP impossible (no tensors in the file), and the measured geometry that priced the shape."
 tags:
  - llm
  - frank-secondary
  - dankclaw
  - llama-cpp
  - gguf
  - rtx3090
 timestamp: 2026-10-07
 ---
 # frank-secondary — llama.cpp Qwen3.5-9B Q6_K, 2 × 200,704, f16 KV
 Live since 2026-10-07, replacing the
 [[frank-secondary-llama-gemma4-q4km-200704]] era (Q4_K_M on ONE 200,704
 slot). The swap is a capacity trade: the 9B dense-hybrid rung costs ~7 GiB of
 weights where the 26B MoE cost ~16.6 GiB, and the freed room buys the second
 conversation slot and unquantized KV back.
 ## Identity
 | | |
 |---|---|
 | Host / device | Dankclaw (192.168.0.143), RTX 3090, 24,576 MiB |
 | Endpoint | `http://192.168.0.143:7771` — alias `frank-secondary`, `/v1` OpenAI surface |
 | Runtime | llama.cpp master (0.5.0-dev, commit `5fc4f3c`), CUDA sm_86 build `…/llama-master/build-cuda/bin/llama-server` (`GGML_CUDA_FA_ALL_QUANTS=ON`); `qwen35` is registered in the build's own `src/llama-arch.cpp` and `src/models/qwen35.cpp` exists at that commit |
 | Unit | root systemd `frank-secondary.service` ([`scripts/frank-secondary.service`](../../scripts/frank-secondary.service)), `User=dank`, `Restart=on-failure`, `LimitMEMLOCK=infinity` |
 | Serve script | [`scripts/serve-frank-secondary.sh`](../../scripts/serve-frank-secondary.sh), tunables [`scripts/frank-secondary.env`](../../scripts/frank-secondary.env) |
 | Weights | `/home/dank/models/Qwen35-9B-HauhauCS-Q6K/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf` — HauhauCS **Q6_K**, 7,359,259,008 B, sha256 `c0ba7beb68fd3fe47891bd549486d38dcf62d00817296ea314ad37017f5a4986` |
 | Projector | `mmproj-Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf`, 921,704,448 B, sha256 `05f662501f8bd45607b079723a3e238a4e888fd085a10a53f4057a0e250f6934` — the model is natively multimodal and the Q6_K file carries no vision tensors, so this file IS the vision capability |
 | Fetch | [`scripts/download-frank-secondary.sh`](../../scripts/download-frank-secondary.sh) — `-f`, byte-count + sha256 verified (the hashes are computed locally: the repo is Xet-backed, so the Hub publishes file OIDs, not file sha256), idempotent, `--verify-only` |
 | Seam | `tests/frank_secondary_unit.bats` (argv assertions against a stub engine) |
 ## Payload geometry (read from the GGUF header, not the model card)
 Range-read of the file's own metadata + tensor-info section:
 |Key|Value|
 |---|---|
 |`general.architecture`|`qwen35` (dense hybrid: Gated DeltaNet + softmax attention)|
 |`qwen35.block_count`|32, `full_attention_interval` 4 ⇒ **8 full-attention layers (3, 7, 11, …, 31) + 24 DeltaNet layers**|
 |attention|16 heads, **4 KV heads**, `key_length` = `value_length` = 256|
 |`qwen35.ssm.*`|`conv_kernel` 4, `state_size` 128, `group_count` 16, `time_step_rank` 32, `inner_size` 4096|
 |context / vocab|262,144 native / 248,320|
 |tensor types|Q6_K × 250 + F32 × 177 — standard types only, so the stock master build loads it (the ROCmFPX-style custom-quant lock-in does not apply)|
 |MTP|**no `nextn`/`mtp` tensors exist** — in-file MTP is impossible; the card's "MTP support" line describes the base architecture, not this quantized export|
 |chat template|supports `enable_thinking`, `tools`/`tool_call`, image/video content parts; it has **no `reasoning_effort` variable** (unused Jinja variables are ignored, so the harness's effort map is inert here, not an error source)|
 ## Shape
 `--ctx-size 401408 --parallel 2` — llama.cpp divides the pool across slots,
 so each conversation gets `n_ctx_slot = 200,704` (Frank's cap; the harness
 `contextWindow` quotes the per-slot number). `--flash-attn on`,
 **`--cache-type-k f16 --cache-type-v f16`** (exact KV), batch 4096 /
 **ubatch 2048**, `--log-verbosity 4`, `--jinja --reasoning auto
 --reasoning-format deepseek`, `--mmproj`, **speculative decoding OFF**
 (`SPEC_TYPE=none` strips both spec flags — the file has no MTP tensors).
 **KV rate law.** The pool scales with the 8 full-attention layers: 8 × 4 KV
 heads × 256 dim = 8,192 elements/token per cache, so f16 K + f16 V bills
 32,768 B/token = **32 KiB/token** → 12.25 GiB at the 401,408 pool. The 24
 DeltaNet layers bill a fixed recurrent state **per slot**, not per token —
 that is what makes two 200k slots affordable where the 26B MoE fit only one.
 **Priced from boots, not formulas.** MiMo-V2.6-Distill-Qwen-9B (identical
 qwen35 geometry, 7,793,710,624 B, same mmproj class) at 2 × 200,704 f16 K +
 f16 V measured **21,276 used / 2,849 free** at ubatch 1024 and **21,846 /
 2,279** at ubatch 2048 — see [[frank-secondary-llama-mimo-q6k-2x200704]].
 This file is 415 MiB lighter in weights and its projector is 3 MiB heavier,
 so the projection is ~21,430 used / ~2,700 free.
 Ubatch 2048 is the measured CUDA optimum for this card and arch (+4.6 %
 prefill at 8.9k depth, decode flat, cost 616 MiB); `1024` is the documented
 fallback rung if a future shape lands under the 2 GiB floor.
 ## Measured boot (2026-10-07)
 TO BE FILLED BY THE DEPLOY VERIFICATION.
 ## Client contract
 llama.cpp surface: model id `frank-secondary`, native tools via `--jinja`,
 thinking in **`reasoning_content`**, vision via OpenAI `image_url` content
 parts (probe with a ≥ 256 px image, or the Qwen-VL tower answers "video").
 The publisher's recipe ships in `models.yml` `extraBody`: thinking mode
 `temperature 0.6 / top_p 0.95 / top_k 20 / min_p 0` (non-thinking mode is
 0.7/0.8/20/0). The card's advice "maintain at least 128K context to preserve
 thinking" is satisfied by the 200,704 slot.
 **Compaction hub**: `models.yml` routes `compactionModel` for all three
 endpoints here. With `--parallel 2` a compaction request no longer serializes
 behind a live generation (the Gemma4 era's `--parallel 1` law), and the
 200,704 slot still fits the worst-case compaction input (~133k tokens).
 ## Rollback
 |Target|How|
 |---|---|
 |Gemma4-26B-A4B Q4_K_M, 1 × 200,704, f16 K + q8_0 V|revert the swap commit — payload stays at `/home/dank/models/Gemma4-26B-A4B/`; see [[frank-secondary-llama-gemma4-q4km-200704]]|
 |MiMo-V2.6 Q6_K, 2 × 200,704, f16 KV|`MODEL_FILE=/home/dank/models/MiMo-Q6K/MiMo-V2.6-Distill-Qwen-9B-Q6_K.gguf` + its mmproj — measured 21,276/2,849 at ubatch 1024, 21,846/2,279 at 2048|
 |Swift-1.5-27B IQ3_XXS 2 × 114,688 + MTP|payload on disk at `/home/dank/models/Swift15-UncIQ3XXS/`|
 ## When it misbehaves
 Same decision tree as the predecessors: `No CUDA GPUs are available` ⇒
 `nvidia-modprobe -u -c=0`; a 400 naming 200,704 ⇒ the conversation outgrew
 one slot, not the pool; queue behind two long generations ⇒ the slot ceiling
 is two, by contract; a boot that OOMs ⇒ read the verbosity-4 ledger in
 `journalctl -u frank-secondary` (KV MiB, compute buffers) before touching
 flags — the engine prints the exact figures the shape needs.
@@ -102,13 +102,20 @@ frank-secondary, port 7771]
  (`slow`/`advisor` → frank-primary) carried forward. Superseded by the
  MODEL swap: the abliterated v2 .hgn loads only in flash-server, so the
  engine moved with the model
- [[frank-secondary-llama-gemma4-q4km-200704]] — **ACTIVE since 2026-10-07**
+- [[frank-secondary-llama-qwen35-9b-q6k-2x200704]] — **ACTIVE since 2026-10-07**
  (root systemd unit, port 7771): llama.cpp master (CUDA sm_86) serving
-  HauhauCS/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-MTP **Q4_K_M**
+  HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive **Q6_K** on the RTX 3090
-  on the RTX 3090 — ONE 200,704-token conversation, q8_0 K + q8_0 V,
+  — TWO 200,704-token conversations (`--ctx-size 401408 --parallel 2`), exact
-  vision via the publisher's mmproj. MTP is OFF (the gemma4-assistant
+  f16 K + f16 V, vision via the publisher's BF16 mmproj. MTP is OFF because
-  drafter costs ~1.9 GiB: 23,806 used/319 free with it, 21,890/2,235
+  the file has no MTP tensors at all. Priced from the MiMo Q6_K boots (same
-  without). Supersedes the Swift-1.5 era (below).
+  qwen35 geometry, same card, same shape). Supersedes the Gemma4 era (below).
 - [[frank-secondary-llama-gemma4-q4km-200704]] — **RETIRED 2026-10-07** (same
  day; superseded by qwen35-9b-q6k; payload on disk as rollback): llama.cpp
  master (CUDA sm_86) serving HauhauCS/Gemma4-26B-A4B-QAT-Uncensored-
  HauhauCS-Balanced-MTP **Q4_K_M** on the RTX 3090 — ONE 200,704-token
  conversation, f16 K + q8_0 V, vision via the publisher's mmproj. MTP was
  OFF (the gemma4-assistant drafter costs ~1.9 GiB: 23,806 used/319 free with
  it, 21,890/2,235 without).
 - [[frank-secondary-llama-swift15-unc-iq3xxs-mtp-114688]] — **RETIRED
  2026-10-07** (superseded by gemma4-q4km; payload on disk as rollback);
  ACTIVE 2026-10-05 (evening) → 2026-10-07; UNCECERSED RCO-IQ3_XXS with
@@ -372,10 +372,16 @@ The LLM inference fleet on the two hosts. Full index:
 - [[ciru-activefpx-promptforge]] — the RETIRED v3 CIRU PromptForge runtime on Dankclaw (superseded 2026-08-24)
 - [[kingjones-strix-mtp]] — the kingjones Qwen3.8-27B ROCmFP4 STRIX MTP model on Dankclaw (active)
 - [[dealign-ornith-crack]] — the retired Ornith-1.5-35B-A3B CRACK model on Dankserver (retired 2026-09-02)
- [[frank-secondary-llama-gemma4-q4km-200704]] — the **ACTIVE** frank-secondary
+- [[frank-secondary-llama-qwen35-9b-q6k-2x200704]] — the **ACTIVE** frank-secondary
-  runbook since 2026-10-07: llama.cpp master (CUDA sm_86) serving
+  runbook since 2026-10-07 (late): llama.cpp master (CUDA sm_86) serving
  HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive **Q6_K** on the RTX 3090 —
  TWO 200,704-token conversations, exact f16 K + f16 V, MTP **impossible**
  (the file carries no `nextn`/`mtp` tensors), vision through the publisher's
  BF16 mmproj; priced from the MiMo Q6_K boots (same qwen35 geometry)
 - [[frank-secondary-llama-gemma4-q4km-200704]] — RETIRED 2026-10-07 (same day;
  payload on disk as rollback): llama.cpp master (CUDA sm_86) serving
  HauhauCS Gemma4-26B-A4B QAT Uncensored **Q4_K_M** on the RTX 3090 — ONE
-  200,704-token conversation, q8_0 K + q8_0 V, MTP **disabled** (the
+  200,704-token conversation, f16 K + q8_0 V, MTP **disabled** (the
  `gemma4-assistant` drafter costs ~1.9 GiB: boot was 23,806 used/319 free
  with it, 21,890/2,235 without), vision through the publisher's mmproj
 - [[frank-secondary-llama-swift15-unc-iq3xxs-mtp-114688]] — RETIRED 2026-10-07
@@ -1,13 +1,13 @@
 # Seam: the frank-secondary llama.cpp root unit on dankclaw.
 # frank-secondary is the RTX 3090 endpoint; since 2026-10-07 it serves
-# HauhauCS's Gemma4-26B-A4B-QAT-Uncensored (QAT Q4_K_M, 30-layer MoE: 25
+# HauhauCS's Qwen3.5-9B-Uncensored-HauhauCS-Aggressive (Q6_K, qwen35 hybrid:
-# sliding-window + 5 full-attention layers) via the CUDA (sm_86) build of
+# 32 layers, 8 full-attention + 24 Gated DeltaNet, no MTP tensors in the file)
-# llama.cpp master: ONE 200,704-token conversation (--parallel 1), f16 K +
+# via the CUDA (sm_86) build of llama.cpp master: TWO 200,704-token
-# q8_0 V under flash attention, the publisher's own BF16 vision projector,
+# conversations (--ctx-size 401408 --parallel 2), exact f16 K + f16 V under
-# and MTP OFF — Gemma4's MTP is a separate assistant file whose KV is built
+# flash attention, the publisher's own BF16 vision projector, and MTP off —
-# against the target's full window (~1.9 GiB measured), which the f16-K raise
+# the file carries no nextn/mtp tensors, so there is no in-file head to draft
-# cannot afford on a 24 GiB card. Log verbosity 4 ships so the VRAM-fit
+# with. Log verbosity 4 ships so the VRAM-fit forensics (KV layout, buffer
-# forensics (KV layout, buffer sizes, graph capture) land in the journal.
+# sizes, graph capture) land in the journal.
 # This seam runs the serve script against a stub engine and asserts on the
 # argv it actually execs, so what is pinned is the command the operator gets.
@@ -34,7 +34,7 @@ STUB
    # (the env) instead of silently forking the test's filename constants.
    MODEL_STUB="$(basename "$(sed -nE 's/^MODEL_FILE=(.+)$/\1/p' "$ENV_FILE")")"
    MM_STUB="$(basename "$(sed -nE 's/^MMPROJ_FILE=(.+)$/\1/p' "$ENV_FILE")")"
-    DRAFT_STUB="mtp-gemma-4-26B-A4B-it.gguf"
+    DRAFT_STUB="draft-assistant.gguf"
    touch "$MODEL_DIR/$MODEL_STUB" "$MODEL_DIR/$MM_STUB" "$MODEL_DIR/$DRAFT_STUB"
 }
@@ -95,10 +95,10 @@ run_stage() {
    grep -q 'FRANK_SECONDARY_ENV_FILE=/home/dank/serverAppConfigs/scripts/frank-secondary.env' "$UNIT"
 }
-@test "the shipped env points at the llama.cpp runtime and the Gemma4-26B-A4B QAT Q4_K_M GGUF" {
+@test "the shipped env points at the llama.cpp runtime and the Qwen3.5-9B Q6_K GGUF" {
    # The payload path IS the endpoint's identity.
    grep -qE '^LLAMA_BIN=/home/dank/GitPackages/llama-master/build-cuda/bin/llama-server$' "$ENV_FILE"
-    grep -qE '^MODEL_FILE=/home/dank/models/Gemma4-26B-A4B/Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M\.gguf$' "$ENV_FILE"
+    grep -qE '^MODEL_FILE=/home/dank/models/Qwen35-9B-HauhauCS-Q6K/Qwen3\.5-9B-Uncensored-HauhauCS-Aggressive-Q6_K\.gguf$' "$ENV_FILE"
 }
@test "the deployer gates on exactly the payload paths the env file ships" {
@@ -115,7 +115,7 @@ run_stage() {
    done
 }
-@test "one conversation owns the whole 200704 window" {
+@test "two conversations share the pool at a 200704 window each" {
    serve_argv
    [ "$status" -eq 0 ]
    [ "$(argv_value --alias)" = "frank-secondary" ]
@@ -124,26 +124,30 @@ run_stage() {
    [ "$(argv_value --model)" = "$MODEL_DIR/$MODEL_STUB" ]
    [ "$(argv_value --n-gpu-layers)" -gt 0 ]
    # llama.cpp divides --ctx-size across slots, so the observable contract
-    # is the per-slot window. The 24 GiB card cannot fund a second 200,704
+    # is the per-slot window. Frank's 200k cap is a per-conversation cap, so
-    # slot at this weight + KV shape (measured: the MTP drafter alone costs
+    # the pair is 401,408 / 2 — and the harness contextWindow must quote
-    # ~1.9 GiB), so --parallel 1 is the shipped contract and the harness
+    # 200704, not the pool and not the model's native 262,144.
-    # contextWindow must quote 200704, not the model's native 262144.
+    [ "$(argv_value --parallel)" -eq 2 ]
-    [ "$(argv_value --parallel)" -eq 1 ]
+    [ "$(argv_value --ctx-size)" -eq 401408 ]
-    [ "$(argv_value --ctx-size)" -eq 200704 ]
+    local pool slots
    pool="$(argv_value --ctx-size)"; slots="$(argv_value --parallel)"
    [ $((pool % slots)) -eq 0 ]
    [ $((pool / slots)) -eq 200704 ]
 }
-@test "the KV contract is f16 K + q8_0 V, flash-attn on" {
+@test "the KV contract is exact f16 K + f16 V, flash-attn on" {
    serve_argv
    [ "$status" -eq 0 ]
-    # Frank, 2026-10-07: K raised to f16 (exact), V stays q8_0 — llama.cpp's
+    # Frank, 2026-10-07: the 9B rung funds unquantized KV (the 26B MoE era
-    # FP8 class (1.0625 B/elem; no true fp8 KV dtype exists). KV scales ONLY
+    # shipped q8_0 V because the card was full). qwen35 bills the pool on the
-    # with the 5 full-attention layers of 30 (the 25 sliding-window layers
+    # 8 full-attention layers only (8 × 4 KV heads × 256 dim = 8,192
-    # cost a fixed 1024-token ring each), so the pair costs ~6.2 GiB at
+    # elements/token ⇒ 32 KiB/token at f16/f16, 12.25 GiB at the 401,408
-    # 200,704 — the boot ledger, not this file, is the referee. flash-attn is
+    # pool); the 24 DeltaNet layers cost a fixed recurrent state per slot.
-    # REQUIRED for quantized V, and the build carries GGML_CUDA_FA_ALL_QUANTS=ON;
+    # The boot ledger, not this file, is the referee. flash-attn stays pinned:
-    # without the flag the engine silently falls back and the contract is a lie.
+    # any quantized-V fallback (q8_0 = this build's FP8 class) requires it, and
    # the build carries GGML_CUDA_FA_ALL_QUANTS=ON.
    [ "$(argv_value --cache-type-k)" = "f16" ]
-    [ "$(argv_value --cache-type-v)" = "q8_0" ]
+    [ "$(argv_value --cache-type-v)" = "f16" ]
    [ "$(argv_value --flash-attn)" = "on" ]
 }
@@ -190,12 +194,12 @@ run_stage() {
 }
@test "speculation is off: no drafter flag reaches the argv" {
-    # Frank, 2026-10-07: MTP OFF on Gemma4. The assistant is a SEPARATE file
+    # Frank, 2026-10-07: MTP is off because it CANNOT be on — the Q6_K file
-    # and its draft KV is built against the target's full 200,704 window —
+    # carries no nextn/mtp tensors (verified from its tensor-name list), and
-    # ~1.9 GiB measured (23,806/319 free), a knife-edge once K is f16. The
+    # MTP is a property of the file, not of the flags. The flags move as a
-    # flags move as a set: SPEC_TYPE=none strips --spec-type, --spec-draft-n-max
+    # set: SPEC_TYPE=none strips --spec-type, --spec-draft-n-max AND
-    # AND --spec-draft-model together, so a no-spec boot cannot carry half a
+    # --spec-draft-model together, so a no-spec boot cannot carry half a
-    # spec pair.
+    # spec pair (the ROCmFPX-era baseline arm segfaulted at load doing that).
    serve_argv
    [ "$status" -eq 0 ]
    ! grep -qxF -- '--spec-type' <<<"$output"
@@ -205,10 +209,10 @@ run_stage() {
 }
@test "naming a drafter ships the full MTP trio" {
-    # The opt-back-in path: SPEC_TYPE + SPEC_DRAFT_FILE (the separate
+    # The opt-back-in path (e.g. a future Qwen3.5 -mtp build): SPEC_TYPE +
-    # gemma4-assistant gguf) must reach the engine as a complete set — a
+    # SPEC_DRAFT_FILE must reach the engine as a complete set — a draft-mtp
-    # draft-mtp flag without its model file boots a drafter against the main
+    # flag without its model file boots a drafter against the main weights and
-    # weights and drafts nothing on this architecture.
+    # drafts nothing.
    serve_argv
    {
        echo 'SPEC_TYPE=draft-mtp'
@@ -222,8 +226,9 @@ run_stage() {
 }
@test "vision ships by default and MMPROJ_FILE empty opts out of the image path" {
-    # The Gemma4 GGUF carries no vision tensors, so the projector argument IS
+    # Qwen3.5-9B is natively multimodal, but the Q6_K file carries no vision
-    # the vision capability: no flag, no eyes.
+    # tensors — the publisher's projector argument IS the vision capability:
    # no flag, no eyes.
    serve_argv
    [ "$status" -eq 0 ]
    [ "$(argv_value --mmproj)" = "$MODEL_DIR/$MM_STUB" ]