Local AI Registry

Recipe

qwen3-5-27b-mq4-rx-7900-xtx-24gb-hipfire-tp1

qwen3-5-27b-mq4-rx-7900-xtx-24gb-hipfire-tp1
JSON APIFind compatible recipes

Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.

Record

Status
candidate
Source
localmaxxing
Engine
hipfire
Engine version
0.1.20-alpha+71896daa
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/Qwen3.5-27B
Repository
Qwen/Qwen3.5-27B
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

Source
https://www.localmaxxing.com/en/runs/cmp8fozao00xqo401nzsa1w68

Observed source tokens

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

  1. ./target/release/examples/dflash_spec_demo
  2. --target
  3. ~/.hipfire/models/qwen3.5-27b.mq4
  4. --draft
  5. ~/.hipfire/models/qwen35-27b-dflash.mq4
  6. --prompt-file
  7. benchmarks/prompts/merge_sort_thinking_off.txt
  8. --max
  9. 256
  10. --temp
  11. 0.0
  12. --no-chatml
  13. --kv-mode
  14. q8
  15. --ctx
  16. 4096
FlagValue
--target~/.hipfire/models/qwen3.5-27b.mq4
--draft~/.hipfire/models/qwen35-27b-dflash.mq4
--prompt-filebenchmarks/prompts/merge_sort_thinking_off.txt
--max256
--temp0.0
--kv-modeq8
--ctx4096

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
14,096—254.4153.9observedqwen3-5-27b-mq4-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
qwen3-5-27b-mq4-rx-7900-xtx-24gb-hipfire-tp1
model instance id
qwen-qwen3-5-27b--mq4
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-5-27b-mq4-rx-7900-xtx-24gb-hipfire-tp1-sweep
status
candidate

capabilities

engine

name
hipfire
version
0.1.20-alpha+71896daa

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 @ 71896daa (feat/tbq, sFWHT KV family + multi-GPU) prompt: merge_sort thinking-OFF (md5=253c7ac50857fe6d0e10fb0d2c5e35c0) benchmarks/prompts/merge_sort_thinking_off.txt (27 input tokens) decode mode: DFlash speculative (adaptive block_size 8..16, mean B=16) drafter: qwen35-27b-dflash.mq4 (z-lab Qwen3.5-DFlash native head, MQ4 weights) target: qwen3.5-27b.mq4 (MQ4 weights, KV q8 filtered: 16/64 layers carry KV) kv_cache: q8 — chosen as winner from {q8,asym3,fwht3,fwht4} sweep (τ identical across all modes at this prompt; q8 has lowest KV-write overhead) prompt_normalize: true (default since 2026-04-26) τ: 13.1818 accept_rate: 0.8788 output: 256 tokens emitted AR baseline (same binary same hardware, --ar-baseline): 44.60 tok/s — DFlash speedup 5.70x Hardware: Sapphire Nitro+ RX 7900 XTX, ROCm 7.2.x, k9lin host GPU coordination: gpu-tcas exclusive lease per cell
localmaxxing · observed command
./target/release/examples/dflash_spec_demo --target ~/.hipfire/models/qwen3.5-27b.mq4 --draft ~/.hipfire/models/qwen35-27b-dflash.mq4 --prompt-file benchmarks/prompts/merge_sort_thinking_off.txt --max 256 --temp 0.0 --no-chatml --kv-mode q8 --ctx 4096
localmaxxing · run id
cmp8fozao00xqo401nzsa1w68
localmaxxing · tokenized · arguments
./target/release/examples/dflash_spec_demo, --target, ~/.hipfire/models/qwen3.5-27b.mq4, --draft, ~/.hipfire/models/qwen35-27b-dflash.mq4, --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 atkindurl
2026-08-30T09:10:02Znormalized-recipewww.localmaxxing.com/en/runs/cmp8fozao00xqo401nzsa1w68 ↗