Recipe

qwen3-8-27b-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp1-o1ragvte

qwen3-8-27b-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp1-o1ragvte

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
4302fb599 + neural.download Qwen3.8 TP1 stack; oneAPI 2026.1.1
Accelerators
1
Tensor parallel
1
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/cmt9m8i0b00z3li01o1ragvte

Observed source tokens

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

  1. repro/qwen38-27b-q4km-tp1-b70/run-server.sh

Environment

VariableValue
BUILD_DIR/path/to/llama.cpp/build-sycl-aot-bmg-g31
MODEL_DIR/path/to/qwen3.8-27b-q4km

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
18,19227.8255.4observedqwen3-8-27b-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp1-o1ragvte-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
intel-arc-pro-b70-32gb
id
qwen3-8-27b-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp1-o1ragvte
model instance id
ggml-org-qwen3-8-27b-gguf--q4-k-m
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-8-27b-q4-k-m-intel-arc-pro-b70-32gb-llama-cpp-tp1-o1ragvte-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
4302fb599 + neural.download Qwen3.8 TP1 stack; oneAPI 2026.1.1

serving

max concurrency
1
max context tokens
8,192
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

metadata

localmaxxing · backend
xpu
localmaxxing · hardware label
Intel Arc Pro B70
localmaxxing · notes
Qwen3.8 27B Q4_K_M target-only on one Intel Arc Pro B70. Fixed 12-prompt cold realistic suite, one request per unique prompt, cache_prompt=false, cached_tokens=0, no draft model or speculation. Primary score is the conventional median over the 99 intervals between generated events 1 and 100; the second fresh run measured 27.813629 tok/s. The lab TP1 fusion ladder improved the 26.05 baseline by 6.8-7.0% with 24/24 oracle-exact outputs. Complete patches, clean-host scripts, model hash, context curves, and commands: https://neural.download/models/qwen38-27b-q4km-tp1-b70.html and https://github.com/steveseguin/b70-optimization-lab/tree/main/repro/qwen38-27b-q4km-tp1-b70.
localmaxxing · observed command
MODEL_DIR=/path/to/qwen3.8-27b-q4km BUILD_DIR=/path/to/llama.cpp/build-sycl-aot-bmg-g31 repro/qwen38-27b-q4km-tp1-b70/run-server.sh
localmaxxing · run id
cmt9m8i0b00z3li01o1ragvte
localmaxxing · tokenized · arguments
repro/qwen38-27b-q4km-tp1-b70/run-server.sh
localmaxxing · tokenized · environment · BUILD DIR
/path/to/llama.cpp/build-sycl-aot-bmg-g31
localmaxxing · tokenized · environment · MODEL DIR
/path/to/qwen3.8-27b-q4km
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/cmt9m8i0b00z3li01o1ragvte