Local AI Registry

Recipe

qwen3-6-27b-q8-0-intel-arc-pro-b70-32gb-llama-cpp-tp1

qwen3-6-27b-q8-0-intel-arc-pro-b70-32gb-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
10298 (15586e2d7) local Intel SYCL VDR2 embedded-MTP build
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/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/cmsn6b0bm0074o001uw5f9kod

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. Qwen3.6-27B-Q8_0.gguf
  4. -dev
  5. SYCL0
  6. -ngl
  7. all
  8. -c
  9. 32768
  10. -np
  11. 1
  12. -b
  13. 1024
  14. -ub
  15. 1024
  16. -t
  17. 8
  18. --threads-http
  19. 6
  20. --poll
  21. 50
  22. -lv
  23. 4
  24. -ctk
  25. f16
  26. -ctv
  27. f16
  28. -fa
  29. on
  30. -fit
  31. on
  32. -fitt
  33. 1024
  34. --spec-type
  35. draft-mtp
  36. --spec-draft-n-max
  37. 3
  38. --spec-draft-n-min
  39. 0
  40. --spec-draft-p-split
  41. 0.10
  42. --spec-draft-p-min
  43. 0.00
  44. --spec-draft-backend-sampling
  45. --spec-draft-device
  46. SYCL0
  47. --spec-draft-ngl
  48. all
  49. --spec-draft-type-k
  50. f16
  51. --spec-draft-type-v
  52. f16
  53. --reasoning
  54. off
  55. --ctx-checkpoints
  56. 0
  57. --cache-ram
  58. 0
  59. --no-cache-idle-slots
  60. --no-context-shift
  61. --slots
  62. --metrics
  63. --jinja
  64. --no-kv-unified
  65. --cont-batching
FlagValue
-mQwen3.6-27B-Q8_0.gguf
-devSYCL0
-nglall
-c32768
-np1
-b1024
-ub1024
-t8
--threads-http6
--poll50
-lv4
-ctkf16
-ctvf16
-faon
-fiton
-fitt1024
--spec-typedraft-mtp
--spec-draft-n-max3
--spec-draft-n-min0
--spec-draft-p-split0.10
--spec-draft-p-min0.00
--spec-draft-deviceSYCL0
--spec-draft-nglall
--spec-draft-type-kf16
--spec-draft-type-vf16
--reasoningoff
--ctx-checkpoints0
--cache-ram0

Environment

VariableValue
GGML_SYCL_ENABLE_DNN0
GGML_SYCL_ENABLE_FLASH_ATTN1
GGML_SYCL_ENABLE_GRAPH0
GGML_SYCL_ENABLE_MKL_FA1
GGML_SYCL_ENABLE_OPT1
GGML_SYCL_ENABLE_VMM1
GGML_SYCL_FA_ONEDNN1
GGML_SYCL_FA_ONEDNN_MAX_KV0
GGML_SYCL_GRAPH_CACHE_SIZE0
ONEAPI_DEVICE_SELECTORlevel_zero:*
UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS1
ZES_ENABLE_SYSMAN1
ZE_AFFINITY_MASK0

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
132,768—36785.5observedqwen3-6-27b-q8-0-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-q8-0-intel-arc-pro-b70-32gb-llama-cpp-tp1
model instance id
unsloth-qwen3-6-27b-mtp-gguf--q8-0
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-6-27b-q8-0-intel-arc-pro-b70-32gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
10298 (15586e2d7) local Intel SYCL VDR2 embedded-MTP build

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
2026-09-01T10:20:39Z
serving.max concurrency · reason explicit-source-server-capacity
serving.max concurrency · state known
serving.max context tokens · provenance · captured at
2026-09-01T10:20:39Z
serving.max context tokens · reason explicit-source-context-limit
serving.max context tokens · state known

metadata

localmaxxing · backend
xpu
localmaxxing · batch size
1
localmaxxing · hardware label
Intel Arc Pro B70
localmaxxing · notes
Qwen3.6 27B integrated publisher-MTP Q8_0 on one Intel Arc Pro B70. The fixed 12-prompt realistic suite used one cold request per prompt, cached_tokens=0 throughout, no prompt/KV/context/response/history reuse, and target-verified embedded MTP3. The primary score is the conventional median over 99 inter-token intervals between generated events 1 and 100. Candidate token IDs and content exactly match a fresh same-integrated-artifact non-spec control for all prompts. Eleven rows reached the 512-token request cap; customer-email ended normally at EOS after 248 tokens, after the required generated-token 1/100 timing endpoints. The immutable measured source retains FAIL from an incompatible legacy 4K/128 oracle; the separately sealed matched-control supplement is PASS.
localmaxxing · observed command
ONEAPI_DEVICE_SELECTOR=level_zero:* ZE_AFFINITY_MASK=0 ZES_ENABLE_SYSMAN=1 UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1 GGML_SYCL_ENABLE_VMM=1 GGML_SYCL_ENABLE_GRAPH=0 GGML_SYCL_GRAPH_CACHE_SIZE=0 GGML_SYCL_ENABLE_DNN=0 GGML_SYCL_ENABLE_OPT=1 GGML_SYCL_FA_ONEDNN=1 GGML_SYCL_FA_ONEDNN_MAX_KV=0 GGML_SYCL_ENABLE_MKL_FA=1 GGML_SYCL_ENABLE_FLASH_ATTN=1 llama-server -m Qwen3.6-27B-Q8_0.gguf -dev SYCL0 -ngl all -c 32768 -np 1 -b 1024 -ub 1024 -t 8 --threads-http 6 --poll 50 -lv 4 -ctk f16 -ctv f16 -fa on -fit on -fitt 1024 --spec-type draft-mtp --spec-draft-n-max 3 --spec-draft-n-min 0 --spec-draft-p-split 0.10 --spec-draft-p-min 0.00 --spec-draft-backend-sampling --spec-draft-device SYCL0 --spec-draft-ngl all --spec-draft-type-k f16 --spec-draft-type-v f16 --reasoning off --ctx-checkpoints 0 --cache-ram 0 --no-cache-idle-slots --no-context-shift --slots --metrics --jinja --no-kv-unified --cont-batching
localmaxxing · run id
cmsn6b0bm0074o001uw5f9kod
localmaxxing · tokenized · arguments
llama-server, -m, Qwen3.6-27B-Q8_0.gguf, -dev, SYCL0, -ngl, all, -c, 32768, -np, 1, -b, 1024, -ub, 1024, -t, 8, --threads-http, 6, --poll, 50, -lv, 4, -ctk, f16, -ctv, f16, -fa, on, -fit, on, -fitt, 1024, --spec-type, draft-mtp, --spec-draft-n-max, 3, --spec-draft-n-min, 0, --spec-draft-p-split, 0.10, --spec-draft-p-min, 0.00, --spec-draft-backend-sampling, --spec-draft-device, SYCL0, --spec-draft-ngl, all, --spec-draft-type-k, f16, --spec-draft-type-v, f16, --reasoning, off, --ctx-checkpoints, 0, --cache-ram, 0, --no-cache-idle-slots, --no-context-shift, --slots, --metrics, --jinja, --no-kv-unified, --cont-batching
localmaxxing · tokenized · environment · GGML SYCL ENABLE DNN
0
localmaxxing · tokenized · environment · GGML SYCL ENABLE FLASH ATTN
1
localmaxxing · tokenized · environment · GGML SYCL ENABLE GRAPH
0
localmaxxing · tokenized · environment · GGML SYCL ENABLE MKL FA
1
localmaxxing · tokenized · environment · GGML SYCL ENABLE OPT
1
localmaxxing · tokenized · environment · GGML SYCL ENABLE VMM
1
localmaxxing · tokenized · environment · GGML SYCL FA ONEDNN
1
localmaxxing · tokenized · environment · GGML SYCL FA ONEDNN MAX KV
0
localmaxxing · tokenized · environment · GGML SYCL GRAPH CACHE SIZE
0
localmaxxing · tokenized · environment · ONEAPI DEVICE SELECTOR
level_zero:*
localmaxxing · tokenized · environment · UR L0 ENABLE RELAXED ALLOCATION LIMITS
1
localmaxxing · tokenized · environment · ZES ENABLE SYSMAN
1
localmaxxing · tokenized · environment · ZE AFFINITY MASK
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/cmsn6b0bm0074o001uw5f9kod ↗