Local AI Registry

Recipe

qwen3-8-27b-q4-k-xl-rx-7900-xtx-24gb-llama-cpp-tp1

qwen3-8-27b-q4-k-xl-rx-7900-xtx-24gb-llama-cpp-tp1
JSON APIFind compatible recipes

Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
9731ad3
Accelerators
1
Tensor parallel
1
Context tokens
32,768
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/Qwen/Qwen3.8-27B
Repository
Qwen/Qwen3.8-27B
Status
known
Link type
Exact Hub repository

Public Hugging Face repository confirmed by the Hub API.

Candidate: useful compatibility or speed evidence. The registry does not offer Run until promotion requirements are met.

Observed configuration

llama.cpp

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source
https://www.localmaxxing.com/en/runs/cmt00amga0eeums015vw8ttp3

Observed source tokens

Mechanical split of the source command. Unverified against the engine CLI. The registry does not offer Run for this recipe.

  1. llama-server
  2. -m
  3. Qwen3.8-27B-UD-Q4_K_XL.gguf
  4. --device
  5. Vulkan0
  6. -c
  7. 32768
  8. -ngl
  9. 999
  10. -fa
  11. on
  12. --cache-type-k
  13. q4_0
  14. --cache-type-v
  15. q4_0
  16. --spec-type
  17. draft-mtp
  18. --spec-draft-n-max
  19. 3
  20. --spec-draft-type-k
  21. q4_0
  22. --spec-draft-type-v
  23. q4_0
  24. --parallel
  25. 1
  26. --no-mmproj
  27. --jinja
  28. --reasoning-budget
  29. 0
  30. --alias
  31. qwen38-27b-src
FlagValue
-mQwen3.8-27B-UD-Q4_K_XL.gguf
--deviceVulkan0
-c32768
-ngl999
-faon
--cache-type-kq4_0
--cache-type-vq4_0
--spec-typedraft-mtp
--spec-draft-n-max3
--spec-draft-type-kq4_0
--spec-draft-type-vq4_0
--parallel1
--reasoning-budget0
--aliasqwen38-27b-src

Environment

VariableValue
GGML_VK_ALLOW_GRAPHICS_QUEUE1
GGML_VK_VISIBLE_DEVICES0

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
132,768326.590.7101.7observedqwen3-8-27b-q4-k-xl-rx-7900-xtx-24gb-llama-cpp-tp1-sweep

Remaining fields

Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.

hardware count
1
hardware id
rx-7900-xtx-24gb
id
qwen3-8-27b-q4-k-xl-rx-7900-xtx-24gb-llama-cpp-tp1
model instance id
qwen-qwen3-8-27b--q4-k-xl
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-8-27b-q4-k-xl-rx-7900-xtx-24gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
9731ad3

serving

max concurrency
1
max context tokens
32,768
tensor parallel
1
Provenance & metadata (3)

facts

capabilities.chat · provenance · captured at
2026-08-30T09:10:02Z
capabilities.chat · reason capability-not-verified
capabilities.chat · state unknown
capabilities.reasoning · provenance · captured at
2026-08-30T09:10:02Z
capabilities.reasoning · reason capability-not-verified
capabilities.reasoning · state unknown
capabilities.tools · provenance · captured at
2026-08-30T09:10:02Z
capabilities.tools · reason capability-not-verified
capabilities.tools · state unknown
capabilities.vision · provenance · captured at
2026-08-30T09:10:02Z
capabilities.vision · reason capability-not-verified
capabilities.vision · state unknown
engine.graph mode · provenance · captured at
2026-08-30T09:10:02Z
engine.graph mode · reason runtime-detail-not-published
engine.graph mode · state unknown
serving.kv cache tokens · provenance · captured at
2026-08-30T09:10:02Z
serving.kv cache tokens · reason kv-cache-capacity-not-published
serving.kv cache tokens · state unknown
serving.max concurrency · provenance · captured at
2026-08-31T23:03:15Z
serving.max concurrency · reason server-capacity-derived-from-source-evidence
serving.max concurrency · state known

metadata

localmaxxing · backend
vulkan
localmaxxing · batch size
1
localmaxxing · hardware label
RX 7900 XTX
localmaxxing · notes
Record decode on Windows 11 Pro + official AMD Vulkan ICD (not RADV). llama.cpp master 9731ad3 build-vulkan llama-server, Unsloth Qwen3.8-27B-UD-Q4_K_XL, native MTP n-max 3 (no p-min), -fa on, q4_0 KV, GGML_VK_ALLOW_GRAPHICS_QUEUE=1. Measured with bench_src.py on a 32k slot: code decode mean 90.74 tok/s (3 runs, 11.04 ms/tok), prose 64.99, mixed 77.81, prefill 326.5 tok/s, TTFT 101.7 ms, dedicated VRAM 17.48 GiB. Daily serve is the same stack at 100k (88.4 code / 62.3 prose). 128k and 132k lose hard on code decode and were not submitted.
localmaxxing · observed command
GGML_VK_VISIBLE_DEVICES=0 GGML_VK_ALLOW_GRAPHICS_QUEUE=1 llama-server -m Qwen3.8-27B-UD-Q4_K_XL.gguf --device Vulkan0 -c 32768 -ngl 999 -fa on --cache-type-k q4_0 --cache-type-v q4_0 --spec-type draft-mtp --spec-draft-n-max 3 --spec-draft-type-k q4_0 --spec-draft-type-v q4_0 --parallel 1 --no-mmproj --jinja --reasoning-budget 0 --alias qwen38-27b-src
localmaxxing · run id
cmt00amga0eeums015vw8ttp3
localmaxxing · tokenized · arguments
llama-server, -m, Qwen3.8-27B-UD-Q4_K_XL.gguf, --device, Vulkan0, -c, 32768, -ngl, 999, -fa, on, --cache-type-k, q4_0, --cache-type-v, q4_0, --spec-type, draft-mtp, --spec-draft-n-max, 3, --spec-draft-type-k, q4_0, --spec-draft-type-v, q4_0, --parallel, 1, --no-mmproj, --jinja, --reasoning-budget, 0, --alias, qwen38-27b-src
localmaxxing · tokenized · environment · GGML VK ALLOW GRAPHICS QUEUE
1
localmaxxing · tokenized · environment · GGML VK VISIBLE DEVICES
0
localmaxxing · tokenized · fidelity
faithful

provenance

captured at
2026-08-30T09:10:02Z

sources

captured atkindurl
2026-08-30T09:10:02Znormalized-recipewww.localmaxxing.com/en/runs/cmt00amga0eeums015vw8ttp3 ↗