Home › Evidence › Records › cfg-0140
⚠ This notebook has stopped taking notes: newest record is 23 days old (eng-0280), against an expected cadence of 14 days.

cfg-0140

backfilled
citable URL: https://halobench.com/records/cfg-0140/ — this address never moves; the anchor /records/#cfg-0140 keeps resolving
modelOrnith-1.5-35B-Q4_K_M.gguf · Q4_K_M · rev ca6ea26329c88b78ffd90a85163be2e746c2fafd1024f56db47e499f117f9a7f
engineggml-org/llama.cpp 3653e6d · rocm · host aihydra (igpu)
flagsServing/capability boundary: -ngl 999 -fa on -b 2048 -ub 512 -c 32768 -dev ROCm0 -ctk f16 -ctv f16 -t 16 --load-mode none --parallel 1 --jinja -n 4096. Guard at depth 8000, tau2 airline with openrouter/anthropic/claude-haiku-4.5 simulator, temp 0, seed 42, num_trials 1, max_steps 200, max_errors 50. Bounded throughput cells at d262144 pp1024/tg256 N=5 fresh-process reps under this same artifact/runtime/backend/f16-KV boundary. Model artifact: /home/aihydra/models/ornith-15-35b/Ornith-1.5-35B-Q4_K_M.gguf, 21,713,462,848 bytes, sha256 ca6ea26329c88b78ffd90a85163be2e746c2fafd1024f56db47e499f117f9a7f, source ornith-ai/Ornith-1.5-35B-A3B-GGUF. Tau2 head: 668d3bcd135c02aa3438f987ef45735b7c163ee3.
templateunknown
treeupstream — stock
aged evidence — reconstructed from the archive. HO-011 reviewer-admitted from /home/aihydra/bench-results/ho011-ornith15-q4km-preflight-r6 (screening) and /home/aihydra/bench-results/ho011-ornith15-q4km-full26-r1 (full-26 tau2). The Q4_K_M quant class is measured on its own boundary; it does not inherit Ornith-1.0 Q8_0 or UD-Q4_K_XL capability, does not establish Ornith-1.0/1.5 equivalence, and does not create an MTP/speculation claim. Energy windows are recorded but not joined — this hbreviewer profile lacks Home Assistant credentials.

Capability basis

measured on this config

Runs on this config (6)

rundatekindsuitemetrics
run-05452026-08-21guardguard-c32768-depth8000tasks_passed=4 · tasks_total=4
run-05462026-08-21capabilitytau2-bench-airline@668d3bctasks_passed=4 · tasks_total=5 · mean_score=0.8
run-05472026-08-21performancellama-bench@1prefill_tps=142.379917
run-05482026-08-21performancellama-bench@1decode_tps=20.988162
run-05492026-08-21guardguard-c32768-depth8000tasks_passed=4 · tasks_total=4
run-05502026-08-21capabilitytau2-bench-airline@668d3bctasks_passed=8 · tasks_total=26 · mean_score=0.32