cfg-0168
backfilledfork
citable URL: https://halobench.com/records/cfg-0168/ — this address never moves; the anchor /records/#cfg-0168 keeps resolving
| model | Qwen3.6-35B-A3B-MTP-UD-Q4_K_M.gguf · UD-Q4_K_M · rev 0b21525e972670ed59e1812e170b27c26355381f0656ecc4e25617ece7dac58b |
| engine | ggml-org/llama.cpp 2586f6eddae19bef3dd21a1a0b109cca7bcf1c32 · Vulkan0/Strix-Halo (strix-halo-v0610 fork via RADV, NOT ggml-org ROCm0) · host aihydra (igpu) |
| flags | HO-014 Cell-1 batch/ubatch arm b512/ub256 (f16/f16 KV). llama-bench pp512/tg256 N=5 d131072, -m Qwen3.6-35B-A3B-UD-Q4_K_M.gguf -ngl 999 -fa on -b 512 -ub 256 --load-mode none -dev Vulkan0 -ctk f16 -ctv f16 -t 16 -p 512 -n 256 -d 131072 -r 5, build 2586f6edd Vulkan/RADV, plain mode (-rea off, no MTP). |
| template | not recorded at test time |
| tree | fork — carrying con-0002, con-0009 |
aged evidence — reconstructed from the archive. From HO-014 reviewer-admitted FILL evidence (runner command c1-b512-ub256.command.txt, summary.tsv). batch/ubatch is a perf lever; guard (4/4) clears binary/hazard levers only, no capability claim. Intra-matrix only.
Capability basis
unmeasured — performance numbers on this config stand on a guard alone, not an established capability
Runs on this config (1)
| run | date | kind | suite | metrics |
|---|---|---|---|---|
| run-0609 | 2026-08-23 | performance | ho014a-fill-r1.cell1-latency | decode_tps=35.14 · ttft_ms=1545.5 |
Cited by — computed at build time, never stored
claims clm-0117