loom-fun/bartowski-Qwen3.8-27B-loom

Qwen3.8-27B (Q4_K_M) for AMD Loom

A Loom package (.loom programs and a model.json card) for the unmodified GGUF checkpoint bartowski/Qwen3.8-27B-GGUF (Qwen3.8-27B-Q4_K_M.gguf), generated on an AMD Ryzen AI Max+ 395 "Strix Halo" with the loom-skills workflow.

Quick start

pip install "hrx-loom[hf]"
loom-fetch https://huggingface.co/loom-fun/bartowski-Qwen3.8-27B-loom/resolve/main/model.json --materialize=copy --output=qwen3.8-27b
loom-serve --model=qwen3.8-27b/model.json --device=amdgpu --memory_bytes=52000000000

The server listens on 127.0.0.1:8080 and answers OpenAI-style /v1/chat/completions requests.

Performance vs llama.cpp

metric	llama.cpp	Loom	margin	roofline	Loom % of roofline	DevLab llama.cpp	DevLab Loom	DevLab llama.cpp Vulkan	local ÷ DevLab (llama.cpp, Loom)
decode tok/s, short prompt	20.3	26.2	+29 %	32.3	81 %	21.8	27.0 ¹	21.9	0.93, 0.97
decode tok/s, 2k context	19.3	24.4	+26 %	31.6	77 %	19.3	24.7 ¹	21.0	1.00, 0.99
decode tok/s, 32k context	16.6	19.3	+16 %	28.3	68 %	17.2	19.8 ¹	17.1	0.97, 0.97
prefill tok/s, 1k document	309.0	499.7	+62 %	1185.9	42 %	321.1	459.9 ¹	283.5	0.96, 1.09
prefill tok/s, 2k document	313.1	502.9	+61 %	1181.2	43 %	323.8	458.6 ¹	286.4	0.97, 1.10
prefill tok/s, 8k document	315.0	509.4	+62 %	1154.0	44 %	280.4	452.7 ¹	284.0	1.12, 1.13
prefill tok/s, 16k document	288.9	484.9	+68 %	1119.5	43 %	263.4	418.8 ¹	270.4	1.10, 1.16
prefill tok/s, 32k document	239.1	439.5	+84 %	1056.5	42 %	246.9	370.6 ¹	245.4	0.97, 1.19
4 streams, aggregate decode tok/s	49.0	61.9	+26 %	123.4	50 %	51.7	62.7 ¹	38.9	0.95, 0.99
speculative decoding, drafted tokens accepted	—	76 % ²	—	n/a	n/a	n/a	n/a	n/a	n/a

• llama.cpp: build llamacpp-rocm/b1355, server arguments -ngl 99 -cram 0 --spec-type draft-dflash --spec-draft-model ~/.cache/huggingface/hub/models--z-lab--Qwen3.8-27B-DFlash2-GGUF/snapshots/2d9571f8ce46e151f61c6499c99dee6079e1d610/Qwen3.8-27B-DFlash2-Q4_K_M.gguf --spec-draft-n-max 3 -ngld 99 -c 65536 -kvu -np 4.
• Loom: speculative decoding on (DFlash, depth 3).
• Same GGUF on both engines, tokens per second. Loom: client-side timing; llama.cpp: server-side timings. Runs per value: decode and prefill 2 (mean), 4 streams 3 (median).
• Measured 2026-11-06; device AMD RYZEN AI MAX+ 395 w/ Radeon 8060S, AMD Radeon 8060S Graphics, Ubuntu 26.04; hrx-loom 0.1.0 (hrx-system d01d1c1); source: local run.
• ¹ Measured 2026-11-07; device not recorded; hrx-loom unknown; source: DevLab Dispatch run 42395970032.
• ² Measured 2026-11-07; device AMD RYZEN AI MAX+ 395 w/ Radeon 8060S, AMD Radeon 8060S Graphics, Ubuntu 24.04; hrx-loom 0.1.0 (hrx-system d01d1c1); source: DevLab Dispatch run 42342529061.
• DevLab: the same measurement on an AMD DevLab Dispatch Strix Halo runner; local ÷ DevLab is the local value over the DevLab value of each engine, recorded for comparison, not a verdict.
• DevLab llama.cpp Vulkan: build not recorded, server arguments -ngl 99 -cram 0 --spec-type draft-dflash --spec-draft-model card:parameter/Qwen3.8-27B-DFlash2-Q4_K_M.gguf --spec-draft-n-max 3 -ngld 99 -c 65536 -kvu -np 4; same GGUF, prompts and settings as llama.cpp; measured 2026-11-09 in its own DevLab job (a different runner than the other DevLab columns: compare the engines through ratios, not across jobs); source: DevLab Dispatch run 42580835796. Differs from llama.cpp: decode, decode_2k, decode_32k measured with unified KV, draft, 4 slots: -ngl 99 -cram 0 --spec-type draft-dflash --spec-draft-model card:parameter/Qwen3.8-27B-DFlash2-Q4_K_M.gguf --spec-draft-n-max 3 -ngld 99 -c 65536 -kvu -np 4 (the original job, run 42569157613). Differs from llama.cpp: prefill_1k, prefill_2k, prefill_8k, prefill_16k, prefill_32k measured with 1 slot, no draft: -ngl 99 -cram 0 -c 36864 -np 1. Differs from llama.cpp: streams4 measured with 4 per-slot streams, no draft: -ngl 99 -cram 0 -c 147456 -np 4. Differs from llama.cpp: each metric is the highest of the Vulkan server configurations measured (b12211, separate DevLab jobs); the ROCm baseline's own configuration is the unified pool with drafting. b12211's Vulkan backend has no GPU kernel for the fused flash attention over K/V rows that a unified pool uses (the server logs the CPU fallback); configurations that did not win or did not run: 1 slot, draft.
• Drafted tokens accepted: the share of the drafter's proposed tokens the model accepted during the spec gate's requests.
• Roofline: the hardware ceiling for this GGUF on AMD Ryzen AI Max+ 395 "Strix Halo" (gfx1151) at 216 GB/s and 59 TFLOP/s, from loom-roofline roofline.py (loom-skills b086d86); decode and 4 streams with speculative decoding: verify passes × (1 + acceptance × depth 3) tokens per pass, the drafter's cost not counted; acceptance measured in the row's own session: decode 50 %, decode at 2k 49 %, decode at 32k 49 %, 4 streams 47 %.
• — not measured; n/a not applicable.

Features

feature	status	detail
API	✅	OpenAI-style /v1/chat/completions
tool calling	✅	OpenAI-style tools and tool_calls
reasoning / thinking	✅	on by default; chat_template_kwargs: {"enable_thinking": false} turns it off
reasoning effort	✅	reasoning_effort: low, medium, xhigh (default xhigh)
speculative decoding	✅	DFlash, depth 3, on by default; drafter z-lab/Qwen3.8-27B-DFlash2-GGUF
image input	✅	OpenAI-style image_url parts; projector mmproj-Qwen3.8-27B-f16.gguf
audio input / output	—	not applicable
sampling	✅	temperature, top_p, top_k, seed
context size	✅	262,144 tokens per request
concurrency	✅	4 concurrent requests
max output tokens	✅	4,096 (preset max_tokens)

Correctness

gate	description	result
kernel_tests	Unit tests of the generated kernels against reference implementations	✅ pass (66/66) ¹
golden	llama.cpp reference recorded on the same GGUF: greedy tokens and top-5 log-probs, short prompts and 1k-32k documents	✅ pass ²
compare	Greedy output vs the llama.cpp reference, token by token; every divergence is a near-tie	✅ pass ²
vision	Image input vs llama.cpp with the same projector: identical prompt tokens, replies equal up to near-ties	✅ pass ²
consistency	Concurrent requests give the same output as the same requests run alone	✅ pass ²
spec	Speculative decoding on gives exactly the output of speculative decoding off	✅ pass ²
agent	Multi-turn tool-calling loop: exact prompt rendering, history reuse, valid tool calls, output vs llama.cpp	✅ pass ²
agent_think_off	The tool-calling loop with thinking off	✅ pass ³
sampling	temperature, top_k and seed behave as specified; first-token distribution matches llama.cpp	✅ pass ²
needle	Needle-in-a-haystack prompt at up to half the context, compared with llama.cpp; over-capacity prompt refused	✅ pass ³
needle_full	Needle-in-a-haystack prompt at the full context is served	✅ pass ²
decision	Decision answers (choice, score, noul probabilities) vs llama.cpp on the same prompts	— not applicable

• ¹ local run, 2026-11-06; device AMD RYZEN AI MAX+ 395 w/ Radeon 8060S, AMD Radeon 8060S Graphics, Ubuntu 26.04; hrx-loom 0.1.0 (hrx-system d01d1c1, recovered).
• ² DevLab Dispatch run 42315858300, LEVEL=full, 2026-11-07; device AMD RYZEN AI MAX+ 395 w/ Radeon 8060S, AMD Radeon 8060S Graphics, Ubuntu 24.04; hrx-loom 0.1.0 (hrx-system d01d1c1).
• ³ DevLab Dispatch run 42342529061, LEVEL=quick, 2026-11-07; device AMD RYZEN AI MAX+ 395 w/ Radeon 8060S, AMD Radeon 8060S Graphics, Ubuntu 24.04; hrx-loom 0.1.0 (hrx-system d01d1c1).
• Gates are the loom-benchmark correctness gates (gates.sh); a gate without a recorded run shows no evidence.

Files

• model.json / model.json.lock: the card and the pinned source files (sha256 checked by loom-fetch).
• program/*.loom, weights.loom: the generated Loom programs.
• tokenizer.json: the tokenizer.
• llm_config.json: the configuration the programs were generated from.
• info.json: the measurements, gate evidence and versions this card shows; everything else comes from the files above.
• REPORT.md: how the values above were measured: setup, commands and the value of every run.

Versions: card schema loom-card/2; loom-skills af9f2a0 (recovered); hrx-loom 0.1.0 (hrx-system d01d1c1, recovered); card generator 1ed9a3e.

Source files, fetched by loom-fetch from their own repositories:

• checkpoint: bartowski/Qwen3.8-27B-GGUF @ 87410e7, Qwen3.8-27B-Q4_K_M.gguf
• drafter: z-lab/Qwen3.8-27B-DFlash2-GGUF @ 2d9571f, Qwen3.8-27B-DFlash2-Q4_K_M.gguf
• projector: bartowski/Qwen3.8-27B-GGUF @ 87410e7, mmproj-Qwen3.8-27B-f16.gguf