cfg-0027
backfilled
citable URL: https://halobench.com/records/cfg-0027/ — this address never moves; the anchor /records/#cfg-0027 keeps resolving
| model | NVIDIA-Nemotron-3-Super-120B-A12B-UD-IQ4_XS-00001-of-00003.gguf · UD-IQ4_XS |
| engine | ggml-org/llama.cpp 3653e6d6d547ec763317d9ecd0ace334a7e21359 · rocm · host aihydra (igpu) |
| flags | -ngl 999 -fa on -c 32768 --parallel 1 --load-mode none --jinja -rea off -ctk f16 -ctv f16 |
| template | not recorded at test time |
| tree | upstream — stock |
aged evidence — reconstructed from the archive. tau2 serving config for the Nemotron IQ4_XS arms (Aug 10 comparison arm, Stage E2). Flags from powered.sh (on-box) and bench/queues/queue-e-nemotron-quant.sh; build asserted at launch. Memory block, HF revision and chat-template hash not recorded at test time; GGUF sha256 ledgered in audits/environment.lock-aihydra-2026-08-12.json.
Capability basis
measured on this config
Runs on this config (2)
| run | date | kind | suite | metrics |
|---|---|---|---|---|
| run-0102 | 2026-08-10 | capability | tau2-bench-airline@668d3bc | tasks_passed=10 · tasks_total=16 · mean_score=0.625 |
| run-0108 | 2026-08-11 | capability | tau2-bench-airline@668d3bc | tasks_passed=7 · tasks_total=12 · mean_score=0.5833 |
Cited by — computed at build time, never stored
model pages nemotron3-super