Local AI Registry

Recipe

qwen3-6-27b-q8-0-rtx-3090-24gb-llama-cpp-tp2

qwen3-6-27b-q8-0-rtx-3090-24gb-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
am17an mtp-clean b1-ebe4fca
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/unsloth/Qwen3.6-27B
Repository
unsloth/Qwen3.6-27B
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/cmp4tz8yu000aph016qhe54n8

Observed source tokens

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

  1. ~/dev/llama.cpp-mtp/build/bin/llama-server
  2. -m
  3. ~/models/Qwen3.6-27B-MTP-GGUF/Qwen3.6-27B-Q8_0.gguf
  4. --alias
  5. qwen3.6-27b-mtp
  6. -ngl
  7. 999
  8. -fa
  9. on
  10. -sm
  11. layer
  12. -c
  13. 8192
  14. -b
  15. 2048
  16. -ub
  17. 1024
  18. -np
  19. 1
  20. --no-context-shift
  21. --spec-type
  22. mtp
  23. --spec-draft-n-max
  24. 3
  25. --metrics
  26. --slots
  27. --jinja
  28. --reasoning
  29. off
  30. --host
  31. 0.0.0.0
  32. --port
  33. 8000
FlagValue
-m~/models/Qwen3.6-27B-MTP-GGUF/Qwen3.6-27B-Q8_0.gguf
--aliasqwen3.6-27b-mtp
-ngl999
-faon
-smlayer
-c8192
-b2048
-ub1024
-np1
--spec-typemtp
--spec-draft-n-max3
--reasoningoff
--host0.0.0.0
--port8000

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
18,192—45.5804.4observedqwen3-6-27b-q8-0-rtx-3090-24gb-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-3090-24gb
id
qwen3-6-27b-q8-0-rtx-3090-24gb-llama-cpp-tp2
model instance id
unsloth-qwen3-6-27b--q8-0
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-6-27b-q8-0-rtx-3090-24gb-llama-cpp-tp2-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
am17an mtp-clean b1-ebe4fca

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
cuda
localmaxxing · batch size
1
localmaxxing · hardware label
RTX 3090
localmaxxing · notes
Lenovo ThinkStation P620 (TR Pro 5955WX, 64 GB DDR4-3200 ECC), 2x RTX 3090 (1x Gigabyte Waterforce Xtreme 3090 Ti + 1x NVIDIA Founders Edition 3090), both PCIe Gen4 x16, both 250 W power cap, ReBAR enabled (BAR1 32 GB each), no NVLink, layer-split via -sm layer (auto). Engine: am17an/llama.cpp mtp-clean branch (PR #22673, build b1-ebe4fca). Model: unsloth/Qwen3.6-27B-MTP-GGUF Q8_0 + spliced 15-tensor NeXTN MTP head. MTP n=3 acceptance ~97.5% on this bench (5949/6307 drafts accepted aggregated over the 20-prompt run; max_output_tokens/s 84.0). KV cache f16, no quant. Methodology: vllm bench serve, random 512in/512out, 20 prompts, max-concurrency 1, --save-result. Comparison context: same model class on a near-identical card set (3090 Ti + 3090) ran on the workstation pre-migration as vLLM FP8 (44.17 t/s) and llama.cpp Q8 no-spec (19.22 t/s) in the 2026-04-26 leaderboard batch.
localmaxxing · observed command
~/dev/llama.cpp-mtp/build/bin/llama-server -m ~/models/Qwen3.6-27B-MTP-GGUF/Qwen3.6-27B-Q8_0.gguf --alias qwen3.6-27b-mtp -ngl 999 -fa on -sm layer -c 8192 -b 2048 -ub 1024 -np 1 --no-context-shift --spec-type mtp --spec-draft-n-max 3 --metrics --slots --jinja --reasoning off --host 0.0.0.0 --port 8000
localmaxxing · run id
cmp4tz8yu000aph016qhe54n8
localmaxxing · tokenized · arguments
~/dev/llama.cpp-mtp/build/bin/llama-server, -m, ~/models/Qwen3.6-27B-MTP-GGUF/Qwen3.6-27B-Q8_0.gguf, --alias, qwen3.6-27b-mtp, -ngl, 999, -fa, on, -sm, layer, -c, 8192, -b, 2048, -ub, 1024, -np, 1, --no-context-shift, --spec-type, mtp, --spec-draft-n-max, 3, --metrics, --slots, --jinja, --reasoning, off, --host, 0.0.0.0, --port, 8000
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/cmp4tz8yu000aph016qhe54n8 ↗