Strix Halo (gorgon-halo) DwarfStar ROCm 10.0 DeepSeek-V4.1-Flash Q2 depth-0 prefill

Protocol deterministic-prime-extend-completions-v1, runner local-llm-speed v1, measured at the OpenAI-completions API endpoint. Mean of 2 repeats after 1 warmup. Concurrency 1, so the figure is per-request, not aggregate. Depth is the primed KV prefix; at depth 0 the whole prompt is computed (actual cached depth 0, recomputed history 0). Prefill is effective follow-up prompt processing: uncached tokens divided by client-observed time to first output. Decode is client-observed generation throughput over the tokens after the first, so TTFT is excluded. KV cache type default, speculative decoding off. Source: local-llm-benchmarks.dev/data/results.json. Run run-6fc53c1202e61fc0, completed 2026-09-30. Engine DwarfStar 51ed19af418bc13fc825887198799533493916672af78010bbb2642ec5fa6211 /usr/local/bin/ds4-server, runtime ROCm 10.0. Container docker.io/kyuz0/strix-halo-ds4-toolbox:rocm-10.0. Weights: antirez/deepseek-v4.1-flash-gguf file DeepSeek-V4.1-Flash-Q2.gguf, quant Q2. Prompt 2048 tokens, 256 generated, primed depth 0.

Host hardware_id is gorgon-halo, a second Strix Halo box in the same dataset, not the Framework Desktop. Q2 main weights are 151.77 GiB plus 188.83 GiB of Engram tables streamed from disk, so this is not a fully resident run.