Recipe
qwen3-6-27b-mq4-awq-rx-7900-xtx-24gb-hipfire-tp1
qwen3-6-27b-mq4-awq-rx-7900-xtx-24gb-hipfire-tp1Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- hipfire
- Engine version
- 0.2.0+ca30ca21
- 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/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
hipfire
Evidence only · candidate · reference
Candidate evidence — not a Run contract
Observed source tokens
Mechanical split of the source command. Unverified against the engine CLI. The registry does not offer Run for this recipe.
./target/release/examples/dflash_spec_demo--target/home/kaden/.hipfire/models/qwen3.6-27b-awq.mq4--draft/home/kaden/.hipfire/models/qwen36-27b-dflash-mq4.hf4--prompt-filebenchmarks/prompts/merge_sort_thinking_off.txt--max256--temp0.0--no-chatml--kv-modeq8--ctx4096
| Flag | Value |
|---|---|
--target | /home/kaden/.hipfire/models/qwen3.6-27b-awq.mq4 |
--draft | /home/kaden/.hipfire/models/qwen36-27b-dflash-mq4.hf4 |
--prompt-file | benchmarks/prompts/merge_sort_thinking_off.txt |
--max | 256 |
--temp | 0.0 |
--kv-mode | q8 |
--ctx | 4096 |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 4,096 | 334.6 | 232.1 | 154.2 | observed | qwen3-6-27b-mq4-awq-rx-7900-xtx-24gb-hipfire-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
- rx-7900-xtx-24gb
- id
- qwen3-6-27b-mq4-awq-rx-7900-xtx-24gb-hipfire-tp1
- model instance id
- qwen-qwen3-6-27b--mq4-awq
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- qwen3-6-27b-mq4-awq-rx-7900-xtx-24gb-hipfire-tp1-sweep
- status
- candidate
capabilities
engine
- name
- hipfire
- version
- 0.2.0+ca30ca21
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
metadata
- localmaxxing · backend
- rocm
- localmaxxing · hardware label
- RX 7900 XTX
- localmaxxing · notes
- hipfire @ ca30ca21 (dense-dflash-gfx11-perfmaxx, origin/feat/dense-dflash-perfmaxx base) prompt: merge_sort thinking-OFF (md5=253c7ac50857fe6d0e10fb0d2c5e35c0) decode mode: DFlash speculative, adaptive B stayed at B=16 target: qwen3.6-27b-awq.mq4 (md5=e42a489edf0be1b144307f2869b73b93) -- corrected AWQ trunk draft: qwen36-27b-dflash-mq4.hf4 (md5=204c4c4ceab30cb9ebc118fa9d59a446) binary: dflash_spec_demo md5=5c7585b5523a6abc0ce9829413a81bec kv_cache: q8; ctx=4096; temp=0.0; no-chatml; prompt_normalize: true/default runs: 3 fresh processes; per-run tok/s: [232.28, 231.81, 232.14]; median=232.14; spread=0.20% median details: cycles=13 committed=174 accepted=148 tau=11.3846 accept_rate=0.7590 prefill median: 80.7ms / 334.59 tok/s; ttft median: 154.23ms; peak VRAM: 17888 MB natural EOS at 162 emitted tokens; decoded output eyeballed as fluent merge_sort code This row supersedes invalid scratch runs that used qwen3.6-27b-dev/qwen3.6-27b.mq4-awq-gptq-f2-lmhead-a100.hfq as trunk. No AR baseline rerun in this correction pass. Hardware: Sapphire Nitro+ RX 7900 XTX / gfx1100 on k9lin, HIP 7.2, Ubuntu 24.04.
- localmaxxing · observed command
- ./target/release/examples/dflash_spec_demo --target /home/kaden/.hipfire/models/qwen3.6-27b-awq.mq4 --draft /home/kaden/.hipfire/models/qwen36-27b-dflash-mq4.hf4 --prompt-file benchmarks/prompts/merge_sort_thinking_off.txt --max 256 --temp 0.0 --no-chatml --kv-mode q8 --ctx 4096
- localmaxxing · run id
- cmppe779e002mpl014npvq9st
- localmaxxing · tokenized · arguments
- ./target/release/examples/dflash_spec_demo, --target, /home/kaden/.hipfire/models/qwen3.6-27b-awq.mq4, --draft, /home/kaden/.hipfire/models/qwen36-27b-dflash-mq4.hf4, --prompt-file, benchmarks/prompts/merge_sort_thinking_off.txt, --max, 256, --temp, 0.0, --no-chatml, --kv-mode, q8, --ctx, 4096
- localmaxxing · tokenized · fidelity
- faithful
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/cmppe779e002mpl014npvq9st ↗ |