Qwen2.5-0.5B-Instruct (Q4_K_M) for AMD Loom
A Loom package (.loom programs and a model.json card) for the unmodified GGUF checkpoint bartowski/Qwen2.5-0.5B-Instruct-GGUF (Qwen2.5-0.5B-Instruct-Q4_K_M.gguf), for the AMD Ryzen AI Max+ 395 "Strix Halo" and AMD Radeon AI PRO R9700 GPUs, generated with the loom-skills workflow. Learn more on the Loom ecosystem website.
Quick start
pipx install "hrx-loom[hf]==0.1.0"
Fetch and serve on the AMD Ryzen AI Max+ 395 "Strix Halo" (gfx1151):
loom-fetch https://huggingface.co/loom-fun/bartowski-Qwen2.5-0.5B-Instruct-loom/resolve/main/model.json --preset=chat --materialize=copy --output=Qwen2.5-0.5B-Instruct
loom-serve --model=Qwen2.5-0.5B-Instruct/model.json '--device=amdgpu://?completion_spin_ns=4294967295' --memory_bytes=5000000000
Fetch and serve on the AMD Radeon AI PRO R9700 (gfx1201, 32 GB):
loom-fetch https://huggingface.co/loom-fun/bartowski-Qwen2.5-0.5B-Instruct-loom/resolve/main/model.json --preset=chat-r9700 --materialize=copy --output=Qwen2.5-0.5B-Instruct
loom-serve --model=Qwen2.5-0.5B-Instruct/model.json '--device=amdgpu://?completion_spin_ns=4294967295' --pool_backing=fixed
The server listens on 127.0.0.1:8080 and answers OpenAI-style /v1/chat/completions requests.
On the AMD Radeon AI PRO R9700, the Quick start fetches the chat-r9700 preset of model.json: the default preset with prefill_capacity 4,096 for its 32 GB. The programs, weights and tokenizer are the same.
Performance vs llama.cpp
Strix Halo (gfx1151)
| metric | llama.cpp | Loom | margin | roofline | Loom % of roofline | llama.cpp Vulkan |
|---|---|---|---|---|---|---|
| decode tok/s, short prompt | 293.4 | 311.5 | +6 % | 551.2 | 57 % | 324.1 |
| decode tok/s, 2k context | 279.1 | 273.6 | -2 % | 518.7 | 53 % | 299.8 |
| decode tok/s, 32k context | 167.1 | 170.1 | +2 % | 275.1 | 62 % | 181.0 |
| prefill tok/s, 1k document | 15164.4 | 18074.2 | +19 % | 77768.6 | 23 % | 12499.0 |
| prefill tok/s, 2k document | 15666.3 | 19036.7 | +22 % | 73596.4 | 26 % | 12411.4 |
| prefill tok/s, 8k document | 13328.6 | 15245.4 | +14 % | 55675.2 | 27 % | 8856.4 |
| prefill tok/s, 16k document | 9894.8 | 12420.9 | +26 % | 42029.3 | 30 % | 6004.3 |
| prefill tok/s, 32k document | 6501.2 | 8667.6 | +33 % | 28203.9 | 31 % | 3610.7 |
| 4 streams, aggregate decode tok/s | 872.1 | 1232.3 | +41 % | 2050.5 | 60 % | 943.1 |
| speculative decoding, drafted tokens accepted | n/a | n/a | n/a | n/a | n/a | n/a |
Methodology
- Compared: Loom against llama.cpp ROCm
llamacpp-rocm/b1374, llama.cpp Vulkanllamacpp-vulkan/b12211; the same GGUF, the same DevLab runner and job, the same prompts; Loom served with the active wait. Values are the mean of 3 rounds (4 streams: median of 2); Loom is timed by the client, llama.cpp by its server. - Out of the box on both engines: Loom serves the preset shipped in this package; llama.cpp runs its build defaults plus only the server arguments listed below (offload, context and slots). No tuning sweep on either engine.
- Drafting, 1-stream rows (decode, decode_2k, decode_32k, prefill): no drafting on either engine (no MTP head in the GGUF and no drafter published for this exact model). Each engine is loaded once for all these rows.
- 4 streams: both engines loaded with drafting off, 4 parallel requests, aggregate tokens per second.
- Rows: decode = short-context decode; decode_2k / decode_32k = decode after a ~2k / ~32k-token context; prefill_Nk = prompt processing of an N k-token prompt; streams4 = 4-stream aggregate decode. A margin within +-3 % (the measured same-runner noise) is a tie.
- llama.cpp ROCm server arguments:
1-stream rows: -ngl 99 -cram 0 -c 65536 -np 1; 4 streams: -ngl 99 -cram 0 -c 65536 -kvu -np 4. - llama.cpp Vulkan server arguments:
1-stream rows: -ngl 99 -cram 0 -c 36864 -np 1; 4 streams: -ngl 99 -cram 0 -c 147456 -np 4. - Measured on AMD DevLab Dispatch Strix Halo runners (
devlab-perfjobs); runners differ between sessions. - Measured 2026-11-13; device AMD Strix Halo (DevLab Dispatch runner X2-LA03-S03-D03), Linux; hrx-loom
0.1.0(hrx-systemd01d1c1, recovered); source: DevLab Dispatch run 43392598780. - Strix Halo performance measured on programs
73d0a49; the default path of this package (sha2567042cd0) is identical after canonicalization (fleet/stage_equiv.py at d2be0384d7 against the measured revision 21391f4839; storage, prepare and control differ only inside the logits blocks; checked 2026-11-15). - Strix Halo re-measured on programs
73d0a49on 2026-11-15: within the per-row bands (methodology, 2-4 %) of the published values except streams4 (faster, confirmed in 2 runs; new values shown) (DevLab Dispatch run 43649821117, DevLab Dispatch run 43659277905). - streams4: the card value is the mean of the two re-measure runs' medians (each the median of 3 sessions; streams4 1202.8 and 1261.8); the published value was streams4 1134.6; llama.cpp values are as published (not re-measured).
- llama.cpp Vulkan: measured 2026-11-13; source: DevLab Dispatch run 43392598780. Archive sha256
9759acd. - 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-skillsacd9ac7). - β not measured; n/a not applicable.
Radeon AI PRO R9700 (gfx1201, 32 GB)
| metric | llama.cpp | Loom | margin | roofline | Loom % of roofline | llama.cpp Vulkan |
|---|---|---|---|---|---|---|
| decode tok/s, short prompt | 390.2 | 646.4 | +66 % | 1564.3 | 41 % | 529.4 |
| decode tok/s, 2k context | 364.5 | 516.2 | +42 % | 1472.0 | 35 % | 492.6 |
| decode tok/s, 32k context | 257.2 | 333.5 | +30 % | 780.8 | 43 % | 351.2 |
| prefill tok/s, 1k document | 32118.6 | 44215.7 | +38 % | 282075.9 | 16 % | 26574.5 |
| prefill tok/s, 2k document | 34742.8 | 51005.1 | +47 % | 266943.1 | 19 % | 28701.3 |
| prefill tok/s, 8k document | 31706.8 | 44704.4 | +41 % | 201940.7 | 22 % | 25259.2 |
| prefill tok/s, 16k document | 25408.3 | 36158.4 | +42 % | 152445.4 | 24 % | 19075.0 |
| prefill tok/s, 32k document | 17812.2 | 26021.5 | +46 % | 102298.9 | 25 % | 12534.5 |
| 4 streams, aggregate decode tok/s | 1005.4 | 2273.9 | +126 % | 5819.4 | 39 % | 1040.6 |
| speculative decoding, drafted tokens accepted | n/a | n/a | n/a | n/a | n/a | n/a |
Methodology
- Compared: Loom against llama.cpp ROCm
llamacpp-rocm/b1375, llama.cpp Vulkanllamacpp-vulkan/b12211; the same GGUF, the same DevLab runner and job, the same prompts; Loom served with the active wait. Values are the median of 6 sessions (a session's value is the median of its samples: 3 samples for decode_2k, decode_32k, prefill_8k; one sample for the others) (4 streams: median of 12); Loom is timed by the client, llama.cpp by its server. - Out of the box on both engines: Loom serves the preset shipped in this package; llama.cpp runs its build defaults plus only the server arguments listed below (offload, context and slots). No tuning sweep on either engine.
- Drafting, 1-stream rows (decode, decode_2k, decode_32k, prefill): no drafting on either engine (no MTP head in the GGUF and no drafter published for this exact model). Each engine is loaded once for all these rows.
- 4 streams: both engines loaded with drafting off, 4 parallel requests, aggregate tokens per second.
- Rows: decode = short-context decode; decode_2k / decode_32k = decode after a ~2k / ~32k-token context; prefill_Nk = prompt processing of an N k-token prompt; streams4 = 4-stream aggregate decode. A margin within +-3 % (the measured same-runner noise) is a tie.
- llama.cpp ROCm server arguments:
-ngl 99 -cram 0 -c 65536 -np 1. - llama.cpp Vulkan server arguments:
-ngl 99 -cram 0 -c 36864 -np 1. - Measured on AMD DevLab Dispatch R9700 runners (
devlab-perfjobs); runners differ between sessions. - Measured 2026-11-14; device DevLab Dispatch runner ephemeral-devlab-sm-01-02-r9700s-6-132564945487, AMD Radeon AI PRO R9700S (gfx1201), Linux; hrx-loom
0.1.0(hrx-system commit unknown, recovered); source: DevLab Dispatch run 43633449157. - Radeon AI PRO R9700 performance measured on programs
73d0a49; the default path of this package (sha2567042cd0) is identical after canonicalization (fleet/stage_equiv.py at d2be0384d7 against the measured revision 21391f4839; storage, prepare and control differ only inside the logits blocks; checked 2026-11-15). - llama.cpp Vulkan: measured 2026-11-14; source: DevLab Dispatch run 43633449157. Archive sha256
9759acd. - Roofline: the hardware ceiling for this GGUF on AMD Radeon AI PRO R9700S (gfx1201, 32 GB, 2.35 GHz) at 613 GB/s and 214 TFLOP/s, from loom-roofline
roofline.py(loom-skillsba84a7a). - β not measured; n/a not applicable.
Features
| feature | status | detail |
|---|---|---|
| API | β | OpenAI-style /v1/chat/completions |
| tool calling | β | OpenAI-style tools and tool_calls |
| reasoning / thinking | β | not applicable |
| reasoning effort | β | not applicable |
| speculative decoding | β | not applicable |
| image input | β | not applicable |
| audio input / output | β | not applicable |
| sampling | β | temperature, top_p, top_k, seed |
| context size | β | 32,768 tokens per request |
| concurrency | β | 4 concurrent requests |
| max output tokens | β | 4,096 (preset max_tokens) |
Correctness
| gate | description | Strix Halo | Radeon AI PRO R9700 |
|---|---|---|---|
kernel_tests |
Unit tests of the generated kernels against reference implementations | β pass (56/56) ΒΉ | β pass (56/56) Β² |
golden |
llama.cpp reference recorded on the same GGUF: greedy tokens and top-5 log-probs, short prompts and 1k-32k documents | β pass Β³ | β pass β΄ |
compare |
Greedy output vs the llama.cpp reference, token by token; every divergence is a near-tie | β pass Β³ | β pass β΄ |
vision |
Image input vs llama.cpp with the same projector: identical prompt tokens, replies equal up to near-ties | β not applicable | β not applicable |
consistency |
Concurrent requests give the same output as the same requests run alone | β pass Β³ | β pass β΄ |
spec |
Speculative decoding on gives exactly the output of speculative decoding off | β not applicable | β not applicable |
agent |
Multi-turn tool-calling loop: exact prompt rendering, history reuse, valid tool calls, output vs llama.cpp | β pass Β³ | β pass β΄ |
agent_think_off |
The tool-calling loop with thinking off | β not applicable | β not applicable |
sampling |
temperature, top_k and seed behave as specified; first-token distribution matches llama.cpp | β pass Β³ | β pass β΄ |
logits |
Host logits readback (llm.logits on): at T=0 every returned row's argmax is the device's token, the top-k matches llama.cpp, greedy text is unchanged | β pass (argmax = device token (3 EOG steps); top-5 = llama.cpp's) β΅ | β pass (argmax = device token (3 EOG steps); top-5 = llama.cpp's) βΆ |
decision |
Decision answers (choice, score, noul probabilities) vs llama.cpp on the same prompts | β not applicable | β not applicable |
- ΒΉ DevLab Dispatch run 43659277905, 2026-11-15; device AMD Strix Halo (DevLab Dispatch runner X2-LA03-S01-D01), Linux; hrx-loom
0.1.0(hrx-systemd01d1c1, recovered). - Β² DevLab Dispatch run 43638050231, 2026-11-14; device DevLab Dispatch runner ephemeral-devlab-sm-01-02-r9700s-0-132564945445, AMD Radeon AI PRO R9700S (gfx1201), Linux; hrx-loom
0.1.0(hrx-system commit unknown, recovered). - Β³ DevLab Dispatch run 43659277905, LEVEL=dev, 2026-11-15; device AMD Strix Halo (DevLab Dispatch runner X2-LA03-S01-D01), Linux; hrx-loom
0.1.0(hrx-systemd01d1c1, recovered). - β΄ DevLab Dispatch run 43638050231, LEVEL=dev, 2026-11-14; device DevLab Dispatch runner ephemeral-devlab-sm-01-02-r9700s-0-132564945445, AMD Radeon AI PRO R9700S (gfx1201), Linux; hrx-loom
0.1.0(hrx-system commit unknown, recovered). - β΅ DevLab Dispatch run 43721362070, LEVEL=dev, 2026-11-15; device AMD Strix Halo (DevLab Dispatch runner X2-LA03-S01-D05), Linux; hrx-loom
0.1.0(hrx-systemd01d1c1, recovered). - βΆ DevLab Dispatch run 43721368522, LEVEL=dev, 2026-11-15; device DevLab Dispatch runner ephemeral-devlab-sm-01-02-r9700s-0-132835931069, AMD Radeon AI PRO R9700S (gfx1201), Linux; hrx-loom
0.1.0(hrx-system commit unknown, recovered). - 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 byloom-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/3; loom-skills d2be038; hrx-loom 0.1.0 (install fingerprint 4358bf6e19ba3594); card generator d2be038.
Source files, fetched by loom-fetch from their own repositories:
- checkpoint: bartowski/Qwen2.5-0.5B-Instruct-GGUF @
41ba88d,Qwen2.5-0.5B-Instruct-Q4_K_M.gguf
Model tree for loom-fun/bartowski-Qwen2.5-0.5B-Instruct-loom
Base model
Qwen/Qwen2.5-0.5B