docs: fleet topology update — PP split renamed frank-default-large, NPU endpoint promoted to frank-default-small, 4-model NPU bakeoff matrix (2026-08-16)

docs/FastFlowLM NPU Inference.md

@@ -161,7 +161,7 @@ Proposed role: a small, always-on, low-power endpoint for triage, summarization,
 - Memlock fixed: `/etc/security/limits.d/99-flm-memlock.conf` sets soft and hard memlock to unlimited. A fresh sshd session picks it up via pam_limits. No reboot needed.
 - `gpt-oss:20b` pulled (4/4 files verified, 13.8 GB `model.q4nx` + tokenizer) into `~/.config/flm/models/GPT-OSS-20B-NPU2/`.
 - `flm serve gpt-oss:20b --host 0.0.0.0 -p 52625 --pmode performance` runs as a nohup process. It is not a systemd unit. It does not survive a reboot.
-- The `llama-server` (frank-pp-head) holds an fd on `/dev/accel/accel0` from its own startup (device enumeration, not compute). FLM opened the same device concurrently without conflict. Verified live 2026-08-16.
+- The `llama-server` (frank-default-large, the PP head) holds an fd on `/dev/accel/accel0` from its own startup (device enumeration, not compute). FLM opened the same device concurrently without conflict. Verified live 2026-08-16.
 
 ## Measured Results (2026-08-16, gpt-oss-20b on the Dankclaw NPU)
 
@@ -177,6 +177,30 @@ Decode is flat at 19.3 t/s across all three runs. Output includes `reasoning_con
 
 Long-context prefill is not yet measured (runs 1-3 used under 200 prompt tokens). The 35B-A3B step remains untested.
 
+## Model Bakeoff (measured 2026-08-16, 4 models)
+
+All runs on the Dankclaw NPU (50 TOPS Strix Halo), single stream, `--pmode performance`, PP head (frank-default-large) serving concurrently on 7770. Decode t/s is the FLM-reported rate. Vision accuracy = ground-truth elements out of 4 (text SPECTRUM, red circle left, blue square right, three green triangles bottom).
+
+| Model | Tag | Weights | Decode t/s (short) | 8k prefill t/s | 8k decode t/s | Vision decode t/s | Vision accuracy | Max ctx (registry) |
+|-------|-----|---------|-------------------:|---------------:|--------------:|------------------:|----------------|--------------------|
+| Gemma4-E4B | `gemma4-it:e4b` | 6.8 GB + 912 MB audio | 12.2 | 582 | 9.9 | 11.7 | 4/4 | 32768 |
+| Qwen3.5-9B | `qwen3.5:9b` | 7.3 GB | 7.2 | 343 | 6.6 | 7.3 | 4/4 | 32768 |
+| GPT-OSS-20B | `gpt-oss:20b` | 13.8 GB | 19.3 | 378 | 16.9 | n/a (no vision) | 0/4 | **8192** |
+| Qwen3.6-35B-A3B | `qwen3.6-moe:35b-a3b` | 22.2 GB | 12.7 | 207 | 11.2 | 12.7 | 4/4 | 32768 |
+
+Notes:
+
+- gpt-oss:20b is the fastest decoder (MoE, 3.6B active) but its KV cache fills at ~8k tokens (98.85% occupancy at 7,926; requests past ~8k return 400). It has no vision encoder.
+- 35B-A3B loads 28-29 GiB resident. It works but leaves ~1.5 GiB headroom with the PP head running. Do not run it alongside the head at full load.
+- Vision prefill for the two vision models with ~720 token image input: gemma 141 t/s, qwen3.5 176 t/s, 35B-A3B 68 t/s.
+- All four comply with the same image at 4/4 elements except gpt-oss, which cannot see it.
+
+## Promotion (2026-08-16)
+
+`frank-default-small` now runs the NPU endpoint on Dankclaw (was the Dankserver llama.cpp service). Unit: `scripts/frank-default-small.service` (symlinked into `/etc/systemd/system`), wrapper `scripts/serve-flm-small.sh`, tunables `scripts/frank-default-small.env`. `LimitMEMLOCK=infinity` is set in the unit. Default model: `gemma4-it:e4b`. Swap `FLM_MODEL` in the env to switch.
+
+The PP split was renamed `frank-pp-head` → `frank-default-large` (2026-08-16, commit `48142f5`).
+
 ## Parallel Streams (measured 2026-08-16)
 
 The NPU serializes generations. It runs exactly one stream at a time. Extra streams queue in FIFO order. FLM server flags `--socket` (default 10) and `--q-len` (default 10) cap queue depth only. They do not add concurrency.