Recipe
qwen3-6-27b-q4-k-m-rtx-5090-32gb-llama-cpp-tp1
qwen3-6-27b-q4-k-m-rtx-5090-32gb-llama-cpp-tp1Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- llama.cpp
- Engine version
- Lucebox-Hub dflash 79fe738 (Luce-Org/llama.cpp@luce-dflash fork)
- Accelerators
- 1
- Tensor parallel
- 1
- Context tokens
- 1,024
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/Qwen/Qwen3.6-27B- Repository
- Qwen/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 command could not be tokenized faithfully. The original string stays in metadata. This does not satisfy promotion.
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 1,024 | — | 228.6 | — | observed | qwen3-6-27b-q4-k-m-rtx-5090-32gb-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
- rtx-5090-32gb
- id
- qwen3-6-27b-q4-k-m-rtx-5090-32gb-llama-cpp-tp1
- model instance id
- qwen-qwen3-6-27b--q4-k-m
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- qwen3-6-27b-q4-k-m-rtx-5090-32gb-llama-cpp-tp1-sweep
- status
- candidate
capabilities
engine
- name
- llama.cpp
- version
- Lucebox-Hub dflash 79fe738 (Luce-Org/llama.cpp@luce-dflash fork)
serving
- max context tokens
- 1,024
- 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-not-evidenced
serving.max concurrency · state unknown
metadata
- localmaxxing · backend
- cuda
- localmaxxing · batch size
- 1
- localmaxxing · hardware label
- RTX 5090
- localmaxxing · notes
- Lucebox-Hub test_dflash on RTX 5090 (Blackwell sm_120). 10-run validation: mean 222.64, best 228.56, min 214.33, std 4.06 tok/s. Same merge_sort thinking-off prompt class (25 input tokens, 256 forced output tokens, temp=0) as the hipfire reference. DFlash speculative decoding + DDTree (budget=16) + fast-rollback. Block-diffusion draft from z-lab/Qwen3.6-27B-DFlash (3.46GB BF16). Mean acceptance length 8.83 tokens/step, 55.2% accept rate, 29 draft steps for 256 tokens. AR baseline same hardware: 72.7 tok/s (llama.cpp mainline UD-Q4_K_XL) -> 3.14x speedup.
- localmaxxing · observed command
- DFLASH27B_TOKENIZER=Qwen/Qwen3.6-27B python3 scripts/run.py --raw --prompt "$(cat merge_sort_thinking_off.txt)" --n-gen 256 --target models/Qwen3.6-27B-Q4_K_M.gguf --draft models/draft-3.6 --budget 16 --max-ctx 1024 --fa-window 0
- localmaxxing · run id
- cmph5zm7j00gwpc0104tq6x4m
- localmaxxing · tokenized · fidelity
- lossy
provenance
- captured at
- 2026-08-30T09:10:02Z
sources
| captured at | kind | url |
|---|---|---|
| 2026-08-30T09:10:02Z | normalized-recipe | www.localmaxxing.com/en/runs/cmph5zm7j00gwpc0104tq6x4m ↗ |