Home › Evidence › Records › cfg-0006

cfg-0006

citable URL: https://halobench.com/records/cfg-0006/ — this address never moves; the anchor /records/#cfg-0006 keeps resolving
modelunsloth/Qwen3.5-122B-A10B-MTP-GGUF · UD-Q4_K_M · rev main
engineggml-org/llama.cpp 3653e6d6d547ec763317d9ecd0ace334a7e21359 · rocm · host aihydra (igpu)
flags-ngl 999 -fa on -c 16384 --parallel 1 --load-mode none --jinja -ctk f16 -ctv f16
templatenot recorded at test time
treeupstream — stock

Memory — static estimate vs observed peak

context 16,384 × 1 slot(s) = 16,384 tokens
⌁ total 76.98 GiB of 105 GiB pool · ⌁ headroom 28.02 GiB
⌁ risk utilisation 73.3% — basis estimate (the estimate has never predicted an OOM — clm-0001)

Capability basis

measured on this config

Runs on this config (4)

rundatekindsuitemetrics
run-00052026-08-08guardcapability-guard@v2tool_call_success=1 · empty_arg_calls=0 · tasks_passed=4 · tasks_total=4
run-00062026-08-08performancespec-ab@v1decode_tps=21.97
run-00982026-08-09capabilitytau2-bench-airline@v1.0.1tool_call_success=1 · tasks_passed=3 · tasks_total=3 · mean_score=1
run-00992026-08-09capabilitytau2-bench-airline-nothink@v1.0.1tool_call_success=1 · tasks_passed=5 · tasks_total=5 · mean_score=1

Cited by — computed at build time, never stored