Home › Evidence › Records › cfg-0186
⚠ This notebook has stopped taking notes: newest record is 23 days old (eng-0280), against an expected cadence of 14 days.

cfg-0186

fork
citable URL: https://halobench.com/records/cfg-0186/ — this address never moves; the anchor /records/#cfg-0186 keeps resolving
modelQwen3.8-27B-IQ4_XS-ALL-IMATRIX-Q8-OUT-MTP.gguf · IQ4_XS-ALL-IMATRIX-Q8-OUT · rev ilintar/qwen3.8-27b-gguf-strix-halo (file 16,110,851,680 bytes)
enginepwilkin/llama.cpp d3b5cc43d1fcfce891f2de94d5274ee40eceb21c · rocm · host aihydra (igpu)
flags-dev ROCm0 -ngl 999 -fa on -fit off --load-mode none -c 262144 -b 2048 -ub 512 --parallel 1 --jinja --spec-type draft-dflash --spec-draft-model Qwen3.8-27B-DFlash2-IQ4_XS.gguf --spec-draft-device ROCm0 --spec-draft-ngl 99 --spec-draft-n-min 0 --spec-draft-n-max 3 --spec-draft-p-min 0.10 --metrics --port 5803 (retained-PM4 runtime env as cfg-0185; fronted by slotpin :8090)
samplingtemperature 0 · max_tokens 256
templatenot recorded at test time
treefork — carrying con-0018, con-0019

Capability basis

unmeasured — performance numbers on this config stand on a guard alone, not an established capability

Runs on this config (4)

rundatekindsuitemetrics
run-06682026-09-06performancedecode-depth-served@drive_depth_serveddecode_tps=27.98 · prefill_tps=319
run-06692026-09-06performancedecode-depth-served@drive_depth_serveddecode_tps=23.18 · prefill_tps=198.7
run-06702026-09-06performancedecode-depth-served@drive_depth_serveddecode_tps=22.12 · prefill_tps=127.4
run-06712026-09-06performancedecode-depth-served@drive_depth_serveddecode_tps=17.07 · prefill_tps=83.9