Local AI Registry

Recipe

qwen3-8-27b-ud-q6-k-xl-rtx-4060-ti-16gb-llama-cpp-tp2

qwen3-8-27b-ud-q6-k-xl-rtx-4060-ti-16gb-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
Accelerators
2
Tensor parallel
2
Context tokens
70,000
Max concurrency
4,096
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/unsloth/Qwen3.8-27B-GGUF
Repository
unsloth/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/cmswd7u0108ndms01i45gcbaj

Observed source tokens

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

    1. app/llama-server
    2. --models-preset
    3. /presets/qwen38_models.ini
    4. --models-max
    5. 1
    6. --host
    7. 0.0.0.0
    8. --port
    9. 8001
    10. -lv
    11. 4
    1. hf-repo
    2. =
    3. unsloth/Qwen3.8-27B-GGUF
    1. hf-file
    2. =
    3. Qwen3.8-27B-UD-Q6_K_XL.gguf
    1. split-mode
    2. =
    3. tensor
    1. ctx-size
    2. =
    3. 70000
    1. spec-type
    2. =
    3. draft-mtp
    1. spec-draft-n-max
    2. =
    3. 5
    1. spec-default
    2. =
    3. true
    1. ctx-checkpoints
    2. =
    3. -1
    1. mmap
    2. =
    3. false
    1. flash-attn
    2. =
    3. on
    1. threads
    2. =
    3. 16
    1. parallel
    2. =
    3. 1
    1. batch-size
    2. =
    3. 4096
    1. ubatch-size
    2. =
    3. 1024
    1. jinja
    2. =
    3. true
    1. reasoning
    2. =
    3. on
    1. reasoning-budget
    2. =
    3. 1000
FlagValue
--models-preset/presets/qwen38_models.ini
--models-max1
--host0.0.0.0
--port8001
-lv4

Source notes

[B_Qwen3.8-27B-UD-Q6_K_XL-tensor-MTP3]

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
409670,000934.543.6—observedqwen3-8-27b-ud-q6-k-xl-rtx-4060-ti-16gb-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
rtx-4060-ti-16gb
id
qwen3-8-27b-ud-q6-k-xl-rtx-4060-ti-16gb-llama-cpp-tp2
model instance id
unsloth-qwen3-8-27b-gguf--ud-q6-k-xl
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-8-27b-ud-q6-k-xl-rtx-4060-ti-16gb-llama-cpp-tp2-sweep
status
candidate

capabilities

engine

name
llama.cpp

serving

max concurrency
4,096
max context tokens
70,000
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

metadata

localmaxxing · backend
cuda
localmaxxing · hardware label
RTX 4060 Ti
localmaxxing · notes
[45713] 8.18.389.062 I slot print_timing: id 0 | task 0 | prompt eval time = 5524.68 ms / 5163 tokens ( 1.07 ms per token, 934.53 tokens per second) [45713] 8.18.389.065 I slot print_timing: id 0 | task 0 | eval time = 479796.78 ms / 20921 tokens ( 22.93 ms per token, 43.60 tokens per second) [45713] 8.18.389.065 I slot print_timing: id 0 | task 0 | total time = 485321.46 ms / 26084 tokens [45713] 8.18.389.070 I slot print_timing: id 0 | task 0 | graphs reused = 5148 [45713] 8.18.389.089 I slot print_timing: id 0 | task 0 | draft acceptance = 0.56873 (15640 accepted / 27500 generated), mean len = 3.97
localmaxxing · observed command
app/llama-server \ --models-preset /presets/qwen38_models.ini \ --models-max 1 \ --host 0.0.0.0 \ --port 8001 \ -lv 4 [B_Qwen3.8-27B-UD-Q6_K_XL-tensor-MTP3] hf-repo = unsloth/Qwen3.8-27B-GGUF hf-file = Qwen3.8-27B-UD-Q6_K_XL.gguf split-mode = tensor ctx-size = 70000 spec-type = draft-mtp spec-draft-n-max = 5 spec-default = true ctx-checkpoints = -1 mmap = false flash-attn = on threads = 16 parallel = 1 batch-size = 4096 ubatch-size = 1024 jinja = true reasoning = on reasoning-budget = 1000
localmaxxing · run id
cmswd7u0108ndms01i45gcbaj
localmaxxing · tokenized · arguments
app/llama-server, --models-preset, /presets/qwen38_models.ini, --models-max, 1, --host, 0.0.0.0, --port, 8001, -lv, 4
localmaxxing · tokenized · fidelity
faithful
localmaxxing · tokenized · notes
[B_Qwen3.8-27B-UD-Q6_K_XL-tensor-MTP3]

localmaxxing · tokenized · steps

01234567
app/llama-server--models-preset/presets/qwen38_models.ini--models-max1--host0.0.0.0--port
hf-repo=unsloth/Qwen3.8-27B-GGUFUnknownUnknownUnknownUnknownUnknown
hf-file=Qwen3.8-27B-UD-Q6_K_XL.ggufUnknownUnknownUnknownUnknownUnknown
split-mode=tensorUnknownUnknownUnknownUnknownUnknown
ctx-size=70000UnknownUnknownUnknownUnknownUnknown
spec-type=draft-mtpUnknownUnknownUnknownUnknownUnknown
spec-draft-n-max=5UnknownUnknownUnknownUnknownUnknown
spec-default=trueUnknownUnknownUnknownUnknownUnknown
ctx-checkpoints=-1UnknownUnknownUnknownUnknownUnknown
mmap=falseUnknownUnknownUnknownUnknownUnknown
flash-attn=onUnknownUnknownUnknownUnknownUnknown
threads=16UnknownUnknownUnknownUnknownUnknown
parallel=1UnknownUnknownUnknownUnknownUnknown
batch-size=4096UnknownUnknownUnknownUnknownUnknown
ubatch-size=1024UnknownUnknownUnknownUnknownUnknown
jinja=trueUnknownUnknownUnknownUnknownUnknown
reasoning=onUnknownUnknownUnknownUnknownUnknown
reasoning-budget=1000UnknownUnknownUnknownUnknownUnknown

provenance

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

sources

captured atkindurl
2026-08-30T09:10:02Znormalized-recipewww.localmaxxing.com/en/runs/cmswd7u0108ndms01i45gcbaj ↗