Qwen3.6-35B-A3Bbenchedguard 4/4
②Verdict
A 35-billion-parameter mixture-of-experts model activating 3B per token, held at UD-Q4_K_XL — the reflex-tier case. It decodes at 2.3x the 122B on identical hardware with prefill above 1000 tok/s, and passes the capability guard 4/4 clm-0023. Flash attention is not optional at depth: without it the model loses 88% of its decode from empty context to 131k, against 44% with it clm-0028. Speculation buys almost nothing on varied prompts, and MTP is unavailable because this build carries no NextN layers clm-0026. The community's 121 tok/s headline is four streams pooled on a different quant format — the honest lever on this model is ROCmFP4, worth roughly 27% on the no-speculation floor, at the price of a third fork divergence clm-0026. Agentic capability is unmeasured: the τ² protocol has not been run on this model, and its reflex-tier standing currently rests on throughput plus a guard. A blind eight-task conceptual screen placed it 7/8 on that rubric clm-0004.
③Best configuration
| model | Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf · UD-Q4_K_XL |
| engine | ggml-org/llama.cpp 3653e6d · rocm · host aihydra (igpu) |
| flags | -ngl 999 -fa 1 -b 2048 -ub 512 -ctk f16 -ctv f16 -t 16 |
| template | not recorded at test time |
| tree | upstream — stock |
backfilled aged evidence — reconstructed from the archive · config record cfg-0012
every cell generated from the record at build time · throughput cells from cfg-0012 (same build 3653e6d, fa on, f16 KV) · provenance: measured-here throughout
③aDecode against context depth
④Other configurations tested — each as a delta against best
| variant | Δ decode | Δ turns (paired tasks) | Δ energy | note | records |
|---|---|---|---|---|---|
| flash attention off | -79% @131k | — | — | Invisible at empty context, catastrophic at depth — and quantised KV cannot create a context at all without flash attention. | clm-0028 |
| q8_0 KV · stock build | -35% @131k | — | — | No dequant-patch arm has been run on this model; what the patch recovers here is an open question, not an assumption. | clm-0022 |
| speculation (ngram-mod / ngram-cache) | — | — | — | Close to worthless on varied prompts — the scatter is roughly three times the mean gain, ngram-cache buys nothing, and draft-mtp fails cleanly because this GGUF carries no NextN layers. Measured in the spec A/B harness; the arms are recorded in the claim, not ingested as runs. | clm-0026 |
deltas computed at build time from the named runs, paired-task traces and metered windows — a hazard row shows words where a delta would mislead
⑤Open questions
- τ² under the pinned simulator — never run on this model; the reflex-tier role rests on throughput and a guard until it is clm-0023 clm-0048.
- Wh per correct answer — unmeasured; energy exists only as bench-window totals, not per-answer figures.
- ROCmFP4 requant — the lever that moves the floor rather than the mean clm-0026; a third divergence to carry if the reflex tier matters.
- The tool-grammar ceiling — the first probe measured a context limit and nearly published it as a grammar one; the true ceiling remains unlocated clm-0024.
- The NPU lane publishes this model in FastFlowLM's catalog — a different artifact at a different quantisation on a different runtime; nothing measured here transfers clm-0013.