Home › Evidence › Records › cfg-0036

cfg-0036

backfilled
citable URL: https://halobench.com/records/cfg-0036/ — this address never moves; the anchor /records/#cfg-0036 keeps resolving
modelQwen3.6-35B-A3B-UD-Q4_K_XL.gguf · UD-Q4_K_XL
engineggml-org/llama.cpp 3be50ccc2 · vulkan · 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. Ingested from llama-bench, which measures throughput and not footprint: the memory block, chat-template hash and model revision are absent by construction. Peak memory comes from a Tier 0 capability probe, not from this tool. Everything present here — build commit, backend, KV quant, batch sizes, flash attention — is machine-reported.

Capability basis

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

Runs on this config (10)

rundatekindsuitemetrics
run-01442026-08-12performancellama-bench@10350prefill_tps=1251.48551
run-01452026-08-12performancellama-bench@10350decode_tps=61.83462
run-01462026-08-12performancellama-bench@10350prefill_tps=321.598303
run-01472026-08-12performancellama-bench@10350decode_tps=34.463715
run-01482026-08-12performancellama-bench@10350prefill_tps=705.35212
run-01492026-08-12performancellama-bench@10350decode_tps=51.020735
run-01502026-08-12performancellama-bench@10350prefill_tps=1149.512533
run-01512026-08-12performancellama-bench@10350decode_tps=59.873755
run-01522026-08-12performancellama-bench@10350prefill_tps=507.325549
run-01532026-08-12performancellama-bench@10350decode_tps=43.934256