Home › Evidence › Records › cfg-0104

cfg-0104

backfilled
citable URL: https://halobench.com/records/cfg-0104/ — this address never moves; the anchor /records/#cfg-0104 keeps resolving
modelOrnith-1.0-35B-Q8_0.gguf · Q8_0
engineggml-org/llama.cpp 3653e6d · rocm · host aihydra (igpu)
flags-ngl 999 -fa 1 -b 2048 -ub 512 -ctk f16 -ctv f16 -t 16
templatenot recorded at test time
treeupstream — stock
aged evidence — reconstructed from the archive. Depth-ladder backfill session (2026-08-17), same fingerprint as cfg-0094 (identical llama-bench invocation: build 3653e6d, ROCm, Q8_0, f16 KV, -fa 1 — the `flags` text matches cfg-0094 verbatim so the site's chart series can merge them; --load-mode none was passed on the actual command line, same as cfg-0094's own invocation, it is simply not part of either config's recorded flags string — only depth differs between the two configs' actual runs, not a fingerprint field). A fresh config id is minted per house convention (each backfill invocation gets its own config record). Ingested from llama-bench, which measures throughput and not footprint: the memory block, chat-template hash and model revision are absent by construction. Everything present here — build commit, backend, KV quant, batch sizes, flash attention, load-mode — is machine-reported at launch.

Capability basis

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

Runs on this config (4)

rundatekindsuitemetrics
run-03772026-08-17performancellama-bench@1prefill_tps=173.658178
run-03782026-08-17performancellama-bench@1decode_tps=22.232903
run-03792026-08-17performancellama-bench@1prefill_tps=141.960239
run-03802026-08-17performancellama-bench@1decode_tps=19.36695

Cited by — computed at build time, never stored

model pages ornith-35b