cfg-0082
citable URL: https://halobench.com/records/cfg-0082/ — this address never moves; the anchor /records/#cfg-0082 keeps resolving
| model | DeepSeek-V4-Flash-0731-UD-IQ3_XXS-00001-of-00004.gguf · UD-IQ3_XXS |
| engine | ggml-org/llama.cpp 3653e6d · rocm · host aihydra (igpu) |
| flags | -ngl 999 -fa 1 -b 2048 -ub 512 -ctk f16 -ctv f16 -t 16 --load-mode none |
| template | not recorded at test time |
| tree | upstream — stock |
Capability basis
unmeasured — performance numbers on this config stand on a guard alone, not an established capability
Runs on this config (10)
| run | date | kind | suite | metrics |
|---|---|---|---|---|
| run-0288 | 2026-08-15 | performance | llama-bench@1 | prefill_tps=140.80165 |
| run-0289 | 2026-08-15 | performance | llama-bench@1 | decode_tps=15.29827 |
| run-0290 | 2026-08-15 | performance | llama-bench@1 | prefill_tps=81.27527 |
| run-0291 | 2026-08-15 | performance | llama-bench@1 | decode_tps=12.35019 |
| run-0292 | 2026-08-15 | performance | llama-bench@1 | prefill_tps=58.62236 |
| run-0293 | 2026-08-15 | performance | llama-bench@1 | decode_tps=10.49586 |
| run-0294 | 2026-08-15 | performance | llama-bench@1 | prefill_tps=36.85766 |
| run-0295 | 2026-08-15 | performance | llama-bench@1 | decode_tps=8.8554 |
| run-0296 | 2026-08-15 | performance | llama-bench@1 | prefill_tps=21.5206 |
| run-0297 | 2026-08-15 | performance | llama-bench@1 | decode_tps=6.85544 |
Cited by — computed at build time, never stored
model pages deepseek-v4-flash