Recipe

qwen3-6-27b-ud-q4-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1

qwen3-6-27b-ud-q4-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
9860 (fdb1db877) local SYCL/AOT Qwen3.6 MTP build
Accelerators
1
Tensor parallel
1
Context tokens
4,096
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/unsloth/Qwen3.6-27B-MTP-GGUF
Repository
unsloth/Qwen3.6-27B-MTP-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/cmr6mn5ct0076mn01on3dnpyn

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. on
  3. GPU0/one
  4. B70
  5. with
  6. --parallel
  7. 1
  8. --cache-ram
  9. 0
  10. --ctx-size
  11. 4096
  12. --batch-size
  13. 1024
  14. --ubatch-size
  15. 256
  16. --flash-attn
  17. on
  18. --spec-type
  19. draft-mtp
  20. --spec-draft-n-max
  21. 3
  22. --spec-draft-n-min
  23. 0
  24. --spec-draft-p-min
  25. 0.00
  26. --ctx-checkpoints
  27. 0;
  28. python3
  29. strict
  30. Qwen
  31. realistic
  32. suite
  33. runner
  34. against
  35. http://127.0.0.1:19431
FlagValue
--parallel1
--cache-ram0
--ctx-size4096
--batch-size1024
--ubatch-size256
--flash-attnon
--spec-typedraft-mtp
--spec-draft-n-max3
--spec-draft-n-min0
--spec-draft-p-min0.00
--ctx-checkpoints0;

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
14,09630.7499.8observedqwen3-6-27b-ud-q4-k-xl-intel-arc-pro-b70-32gb-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
intel-arc-pro-b70-32gb
id
qwen3-6-27b-ud-q4-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1
model instance id
unsloth-qwen3-6-27b-mtp-gguf--ud-q4-k-xl
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-6-27b-ud-q4-k-xl-intel-arc-pro-b70-32gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
9860 (fdb1db877) local SYCL/AOT Qwen3.6 MTP build

serving

max concurrency
1
max context tokens
4,096
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
xpu
localmaxxing · batch size
1
localmaxxing · hardware label
Intel Arc Pro B70
localmaxxing · notes
Qwen3.6 27B MTP GGUF Q4 single-B70 llama.cpp/SYCL reference. This is a valid but non-record model-variation result: fixed Qwen realistic cold suite, 12 unique prompts, each prompt once, cached_tokens=0 every row, no prompt/KV/context checkpoint/response reuse, no n-gram/history acceleration, and target-verified draft-MTP speculation. Primary metric is median generated-token throughput for tokens 1-100 after TTFT. This row documents expected performance for the UD-Q4_K_XL GGUF runtime and should not be confused with the faster AutoRound vLLM/XPU result.
localmaxxing · observed command
llama-server on GPU0/one B70 with --parallel 1 --cache-ram 0 --ctx-size 4096 --batch-size 1024 --ubatch-size 256 --flash-attn on --spec-type draft-mtp --spec-draft-n-max 3 --spec-draft-n-min 0 --spec-draft-p-min 0.00 --ctx-checkpoints 0; python3 strict Qwen realistic suite runner against http://127.0.0.1:19431
localmaxxing · run id
cmr6mn5ct0076mn01on3dnpyn
localmaxxing · tokenized · arguments
llama-server, on, GPU0/one, B70, with, --parallel, 1, --cache-ram, 0, --ctx-size, 4096, --batch-size, 1024, --ubatch-size, 256, --flash-attn, on, --spec-type, draft-mtp, --spec-draft-n-max, 3, --spec-draft-n-min, 0, --spec-draft-p-min, 0.00, --ctx-checkpoints, 0;, python3, strict, Qwen, realistic, suite, runner, against, http://127.0.0.1:19431
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/cmr6mn5ct0076mn01on3dnpyn