Home › Evidence › Records › cfg-0188

cfg-0188

citable URL: https://halobench.com/records/cfg-0188/ — this address never moves; the anchor /records/#cfg-0188 keeps resolving
modelQwen3.8-Flash-Next-UD-Q4_K_XL-00001-of-00004.gguf · UD-Q4_K_XL · rev 4448186216b3af4cc558bbce2c3213f01608f8f8b2e5267a9767971dd3ec8082
enginegufo-org/gufo 1b4e6825f2cf6fa2203af3b4b3596bf25e4bfa4e · hip · host aibeast (igpu)
flagsgufo serve --host 127.0.0.1 --port 5830 llm --model Qwen3.8-Flash-Next-UD-Q4_K_XL-00001-of-00004.gguf --served-model-name qwen38-flash-next-ud-q4kxl --speculative mtp --mtp-model mtp-Qwen3.8-Flash-Next-shared-Q8_0.gguf --sessions 1 --context 200192 --reasoning-effort low --max-pending-per-client 16 --cache-disk <dir> --cache-disk-staging-bytes 4294967296. mmproj-BF16 is discovered beside the target. Served through the slotpin energy-metering proxy in gufo mode (upstream_kind gufo; slotpin owns admission, Gufo owns prompt caching). tau2 agent args temperature=0.0, max_tokens=4096, airline seed 42, tasks 0-25, claude-haiku-4.5 simulator, max-steps 200, concurrency 1.
templatenot recorded at test time
treeupstream — stock

Capability basis

measured on this config

production node(s): qwen38-flash-next-gufo

Runs on this config (5)

rundatekindsuitemetrics
run-06742026-10-03capabilitytau2-airline-full26tasks_passed=22 · tasks_total=26
run-06752026-10-03performancedecode-depth-serveddecode_tps=47.5 · prefill_tps=1416
run-06762026-10-03performancedecode-depth-serveddecode_tps=57.6 · prefill_tps=1383
run-06772026-10-03performancedecode-depth-serveddecode_tps=43.7 · prefill_tps=1340
run-06782026-10-03performancedecode-depth-serveddecode_tps=42.5 · prefill_tps=1297

Cited by — computed at build time, never stored

claims clm-0136
model pages qwen38-flash-next