Recipe

qwen3-6-27b-mq4-radeon-ai-pro-r9700-32gb-hipfire-tp3

qwen3-6-27b-mq4-radeon-ai-pro-r9700-32gb-hipfire-tp3

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

Record

Status
candidate
Source
localmaxxing
Engine
hipfire
Engine version
0.2.1+45cb5ab
Accelerators
3
Tensor parallel
3
Context tokens
4,096
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

Source
https://www.localmaxxing.com/en/runs/cmrgp8xgt0062mj01bhjtg8ra

Observed source tokens

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

  1. hipfire
  2. run
  3. qwen3.6:27b
  4. --spec
  5. dflash
  6. -md
  7. qwen36-27b-dflash-mq4.hf4
  8. --temp
  9. 0.0
  10. -n
  11. 256
  12. --kv-mode
  13. q8
FlagValue
--specdflash
-mdqwen36-27b-dflash-mq4.hf4
--temp0.0
-n256
--kv-modeq8

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
4,096750213.763observedqwen3-6-27b-mq4-radeon-ai-pro-r9700-32gb-hipfire-tp3-sweep

Remaining fields

Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.

hardware count
3
hardware id
radeon-ai-pro-r9700-32gb
id
qwen3-6-27b-mq4-radeon-ai-pro-r9700-32gb-hipfire-tp3
model instance id
qwen-qwen3-6-27b--mq4
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-6-27b-mq4-radeon-ai-pro-r9700-32gb-hipfire-tp3-sweep
status
candidate

capabilities

engine

name
hipfire
version
0.2.1+45cb5ab

serving

max context tokens
4,096
tensor parallel
3
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
rocm
localmaxxing · batch size
1
localmaxxing · hardware label
Radeon AI Pro R9700
localmaxxing · notes
BASELINE: DFlash speculative decode on stock hipfire master (0.2.1+45cb5ab), non-AWQ MQ4 target + MQ4 dflash draft, 3x R9700 rig; hipfire runs the 27B on one card (dev 0). Kaden-style prompt merge_sort_thinking_off (md5 253c7ac5), temp 0.0, kv q8, ctx 4096. tokSOut = median of stable warm runs (213.0/213.5/213.7/214.1/214.6); first-run-after-cold-stop ramps to ~141 (clock/graph warmup) and is excluded. Reference points: hipfire MTP MQ4 = 78.8, llama.cpp MTP Q8_0 = 34.7. Target to beat: MQ4-AWQ + perfmaxx build 0.2.0+3730b58 = 254.8.
localmaxxing · observed command
hipfire run qwen3.6:27b --spec dflash -md qwen36-27b-dflash-mq4.hf4 --temp 0.0 -n 256 --kv-mode q8
localmaxxing · run id
cmrgp8xgt0062mj01bhjtg8ra
localmaxxing · tokenized · arguments
hipfire, run, qwen3.6:27b, --spec, dflash, -md, qwen36-27b-dflash-mq4.hf4, --temp, 0.0, -n, 256, --kv-mode, q8
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/cmrgp8xgt0062mj01bhjtg8ra