loom-fun/bartowski-Ornith-1.5-9B-loom

Ornith-1.5-9B (Q4_K_M) for AMD Loom

A Loom package (.loom programs and a model.json card) for the unmodified GGUF checkpoint bartowski/Ornith-1.5-9B-GGUF (Ornith-1.5-9B-Q4_K_M.gguf), generated on an AMD Ryzen AI Max+ 395 "Strix Halo" with the loom-skills workflow. Learn more on the Loom ecosystem website.

Quick start

pip install "hrx-loom[hf]==0.1.0"
loom-fetch https://huggingface.co/loom-fun/bartowski-Ornith-1.5-9B-loom/resolve/main/model.json --materialize=copy --output=Ornith-1.5-9B
loom-serve --model=Ornith-1.5-9B/model.json '--device=amdgpu://?completion_spin_ns=4294967295' --memory_bytes=32000000000

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	llama.cpp Vulkan
decode tok/s, short prompt	49.8	61.7	+24 %	89.4	69 %	34.7
decode tok/s, 2k context	48.5	54.5	+12 %	88.8	61 %	33.8
decode tok/s, 32k context	34.2	42.7	+25 %	68.0	63 %	29.0
prefill tok/s, 1k document	946.4	1610.8	+70 %	4243.3	38 %	859.3
prefill tok/s, 2k document	948.1	1635.9	+73 %	4223.4	39 %	887.6
prefill tok/s, 8k document	754.5	1648.0	+118 %	4107.8	40 %	888.1
prefill tok/s, 16k document	715.5	1587.6	+122 %	3963.1	40 %	844.6
prefill tok/s, 32k document	648.0	1453.4	+124 %	3702.4	39 %	757.5
4 streams, aggregate decode tok/s	113.2	162.1	+43 %	342.5	47 %	108.5
speculative decoding, drafted tokens accepted	—	70 % ¹	—	n/a	n/a	n/a

• llama.cpp: build llamacpp-rocm/b1374, server arguments -ngl 99 -cram 0 --spec-type draft-dflash --spec-draft-model card:parameter/Ornith-1.5-9B-DFlash-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 on AMD DevLab Dispatch Strix Halo runners (devlab-perf jobs): Loom, served with the active wait, and llama.cpp on the same runner; runners differ between sessions.
• Measured 2026-11-13; device AMD Strix Halo (DevLab Dispatch runner X2-LA03-S04-D03), Linux; hrx-loom 0.1.0 (hrx-system d01d1c1, recovered); source: DevLab Dispatch run 43334172733.
• ¹ Measured 2026-11-12; 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 43325632377.
• llama.cpp Vulkan: build llamacpp-vulkan/b12211 (archive sha256 9759acd), server arguments -ngl 99 -cram 0 -c 36864 -np 1; same GGUF, prompts and settings as llama.cpp; measured 2026-11-13 in its own DevLab job (another runner than the other columns: compare the engines through ratios, not across jobs); source: DevLab Dispatch run 43334172733. Differs from llama.cpp: np: 1 slot (36864 tokens) for decode rows, 4 per-slot caches for streams4; no drafting (Vulkan b12211 baseline of a no-draft server).
• 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 14669f7); 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 40 %, decode at 2k 41 %, decode at 32k 34 %, 4 streams 38 %.
• — 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	—	not applicable
speculative decoding	✅	DFlash, depth 3, on by default; drafter abenzerps/Ornith-1.5-9B-DFlash-GGUF
image input	✅	OpenAI-style image_url parts; projector mmproj-Ornith-1.5-9B-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 (72/72) 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 ²
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 ²
needle_full	Needle-in-a-haystack prompt at the full context is served	✅ pass ³
decision	Decision answers (choice, score, noul probabilities) vs llama.cpp on the same prompts	— not applicable

• ¹ DevLab Dispatch run 43325632377, 2026-11-12; device AMD RYZEN AI MAX+ 395 w/ Radeon 8060S, AMD Radeon 8060S Graphics, Ubuntu 24.04; hrx-loom 0.1.0 (hrx-system d01d1c1).
• ² DevLab Dispatch run 43325632377, LEVEL=full, 2026-11-12; 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 42407996692, 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).
• 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 d571bcb; hrx-loom 0.1.0 (hrx-system d01d1c1); card generator d571bcb.

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

• checkpoint: bartowski/Ornith-1.5-9B-GGUF @ 1fc60d6, Ornith-1.5-9B-Q4_K_M.gguf
• drafter: abenzerps/Ornith-1.5-9B-DFlash-GGUF @ 4ba29c7, Ornith-1.5-9B-DFlash-Q4_K_M.gguf
• projector: bartowski/Ornith-1.5-9B-GGUF @ 1fc60d6, mmproj-Ornith-1.5-9B-f16.gguf