Recipe

qwen3-6-27b-q4-k-xl-radeon-ai-pro-r9700-32gb-llama-cpp-tp1-ny0b512u

qwen3-6-27b-q4-k-xl-radeon-ai-pro-r9700-32gb-llama-cpp-tp1-ny0b512u

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Accelerators
1
Tensor parallel
1
Context tokens
65,536
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/unsloth/Qwen3.6-27B-GGUF
Repository
unsloth/Qwen3.6-27B-GGUF
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/cmp99vyrv046co401ny0b512u

Observed source tokens

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

  1. llama-server
  2. -m
  3. Qwen3.6-27B-UD-Q4_K_XL.gguf
  4. --mmproj
  5. mmproj-BF16.gguf
  6. --host
  7. 0.0.0.0
  8. --port
  9. 1234
  10. -c
  11. 65536
  12. -ngl
  13. 999
  14. --flash-attn
  15. on
  16. --jinja
  17. --no-mmap
  18. -fit
  19. off
FlagValue
-mQwen3.6-27B-UD-Q4_K_XL.gguf
--mmprojmmproj-BF16.gguf
--host0.0.0.0
--port1234
-c65536
-ngl999
--flash-attnon
-fitoff

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
165,53694328.3observedqwen3-6-27b-q4-k-xl-radeon-ai-pro-r9700-32gb-llama-cpp-tp1-ny0b512u-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
radeon-ai-pro-r9700-32gb
id
qwen3-6-27b-q4-k-xl-radeon-ai-pro-r9700-32gb-llama-cpp-tp1-ny0b512u
model instance id
unsloth-qwen3-6-27b-gguf--q4-k-xl
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-6-27b-q4-k-xl-radeon-ai-pro-r9700-32gb-llama-cpp-tp1-ny0b512u-sweep
status
candidate

capabilities

engine

name
llama.cpp

serving

max context tokens
65,536
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
rocm
localmaxxing · batch size
1
localmaxxing · hardware label
Radeon AI Pro R9700
localmaxxing · notes
from 2k to 64k depth with llama-benchy, goes from just under 1k prefill down to just under 500. decode goes from 28 to around 23 but didn't test long outputs.
localmaxxing · observed command
llama-server \ -m Qwen3.6-27B-UD-Q4_K_XL.gguf \ --mmproj mmproj-BF16.gguf \ --host 0.0.0.0 \ --port 1234 \ -c 65536 \ -ngl 999 \ --flash-attn on \ --jinja \ --no-mmap \ -fit off
localmaxxing · run id
cmp99vyrv046co401ny0b512u
localmaxxing · tokenized · arguments
llama-server, -m, Qwen3.6-27B-UD-Q4_K_XL.gguf, --mmproj, mmproj-BF16.gguf, --host, 0.0.0.0, --port, 1234, -c, 65536, -ngl, 999, --flash-attn, on, --jinja, --no-mmap, -fit, off
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/cmp99vyrv046co401ny0b512u