Local AI Registry

Recipe

qwen3-8-27b-gguf-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp2

qwen3-8-27b-gguf-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp2
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
4302fb599 + lab TP2 f21e9b557 + Q4K GLU 0a2785852; oneAPI 2026.1.1
Accelerators
2
Tensor parallel
2
Context tokens
8,192
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/ggml-org/Qwen3.8-27B-GGUF
Repository
ggml-org/Qwen3.8-27B-GGUF
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/cmsy530c70cpwms01bl1sjk6g

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-Q4_K_M.gguf
  4. --device
  5. SYCL0,SYCL1
  6. --gpu-layers
  7. 99
  8. --split-mode
  9. tensor
  10. --tensor-split
  11. 1,1
  12. --flash-attn
  13. on
  14. --batch-size
  15. 1024
  16. --ubatch-size
  17. 256
  18. --cache-type-k
  19. f16
  20. --cache-type-v
  21. f16
  22. --cache-ram
  23. 0
  24. --ctx-checkpoints
  25. 0
  26. --fit
  27. off
  28. --reasoning
  29. off
  30. --threads
  31. 8
  32. --poll
  33. 50
  34. --ctx-size
  35. 8192
  36. --parallel
  37. 1
FlagValue
-mQwen3.8-27B-Q4_K_M.gguf
--deviceSYCL0,SYCL1
--gpu-layers99
--split-modetensor
--tensor-split1,1
--flash-attnon
--batch-size1024
--ubatch-size256
--cache-type-kf16
--cache-type-vf16
--cache-ram0
--ctx-checkpoints0
--fitoff
--reasoningoff
--threads8
--poll50
--ctx-size8192
--parallel1

Environment

VariableValue
GGML_SYCL_FUSED_MMVQ_SWIGLU_Q4K1
GGML_SYCL_MMQ_Q4K_REORDER1
ONEAPI_DEVICE_SELECTORlevel_zero:1,0

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
18,192—49.7173.6observedqwen3-8-27b-gguf-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp2-sweep

Remaining fields

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

hardware count
2
hardware id
intel-arc-pro-b70-32gb
id
qwen3-8-27b-gguf-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp2
model instance id
ggml-org-qwen3-8-27b-gguf--gguf-q4-k-m
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-8-27b-gguf-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp2-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
4302fb599 + lab TP2 f21e9b557 + Q4K GLU 0a2785852; oneAPI 2026.1.1

serving

max concurrency
1
max context tokens
8,192
tensor parallel
2
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
xpu
localmaxxing · batch size
1
localmaxxing · hardware label
Intel Arc Pro B70
localmaxxing · notes
Qwen3.8 27B Q4_K_M target-only TP2 on two Intel Arc Pro B70 GPUs. Fixed 12-prompt realistic suite, one cold request per unique prompt, cache_prompt=false and cached_tokens=0 throughout. No MTP, DFlash, draft model, prompt/KV/context/history/response reuse, or speculation. Primary score is the conventional median over 99 inter-token intervals between generated events 1 and 100. Device-local Q4_K gate/up/SwiGLU fusion passed 12/12 exact complete-output hashes against the prior target-only oracle. Public patch and repro are linked in engineFlags. One row ended naturally at EOS after 272 generated tokens, after the required 100-event timing window.
localmaxxing · observed command
ONEAPI_DEVICE_SELECTOR=level_zero:1,0 GGML_SYCL_MMQ_Q4K_REORDER=1 GGML_SYCL_FUSED_MMVQ_SWIGLU_Q4K=1 llama-server -m Qwen3.8-27B-Q4_K_M.gguf --device SYCL0,SYCL1 --gpu-layers 99 --split-mode tensor --tensor-split 1,1 --flash-attn on --batch-size 1024 --ubatch-size 256 --cache-type-k f16 --cache-type-v f16 --cache-ram 0 --ctx-checkpoints 0 --fit off --reasoning off --threads 8 --poll 50 --ctx-size 8192 --parallel 1
localmaxxing · run id
cmsy530c70cpwms01bl1sjk6g
localmaxxing · tokenized · arguments
llama-server, -m, Qwen3.8-27B-Q4_K_M.gguf, --device, SYCL0,SYCL1, --gpu-layers, 99, --split-mode, tensor, --tensor-split, 1,1, --flash-attn, on, --batch-size, 1024, --ubatch-size, 256, --cache-type-k, f16, --cache-type-v, f16, --cache-ram, 0, --ctx-checkpoints, 0, --fit, off, --reasoning, off, --threads, 8, --poll, 50, --ctx-size, 8192, --parallel, 1
localmaxxing · tokenized · environment · GGML SYCL FUSED MMVQ SWIGLU Q4K
1
localmaxxing · tokenized · environment · GGML SYCL MMQ Q4K REORDER
1
localmaxxing · tokenized · environment · ONEAPI DEVICE SELECTOR
level_zero:1,0
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/cmsy530c70cpwms01bl1sjk6g ↗