Recipe

qwen3-8-27b-iq4-xs-ryzen-ai-max-plus-395-128gb-lucebox-tp1

qwen3-8-27b-iq4-xs-ryzen-ai-max-plus-395-128gb-lucebox-tp1

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

Record

Status
candidate
Source
localmaxxing
Engine
lucebox
Engine version
f686c447f067a04ea100a996e4c826e8cc4decc1
Accelerators
1
Tensor parallel
1
Context tokens
131,072
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/unsloth/Qwen3.8-27B-GGUF
Repository
unsloth/Qwen3.8-27B-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

lucebox

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source
https://www.localmaxxing.com/en/runs/cmt7on4h50059nn01bkp53rrp

Observed source tokens

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

  1. /home/soulrider4ever/lucebox-src/server/build-hip-dual/dflash_server
  2. /home/soulrider4ever/lucebox-src/server/models/Qwen3.8-27B-UD-IQ4_XS.gguf
  3. --draft
  4. /home/soulrider4ever/lucebox-src/server/models/draft/qwen38-dflash2-q8_0.gguf
  5. --draft-block-size
  6. 16
  7. --max-ctx
  8. 131072
  9. --cache-type-k
  10. q8_0
  11. --cache-type-v
  12. q8_0
  13. --host
  14. 0.0.0.0
  15. --port
  16. 8087
  17. --target-device
  18. hip:0
  19. --draft-device
  20. hip:0
FlagValue
--draft/home/soulrider4ever/lucebox-src/server/models/draft/qwen38-dflash2-q8_0.gguf
--draft-block-size16
--max-ctx131072
--cache-type-kq8_0
--cache-type-vq8_0
--host0.0.0.0
--port8087
--target-devicehip:0
--draft-devicehip:0

Environment

VariableValue
HIP_VISIBLE_DEVICES1

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
131,072135.396.5368.2observedqwen3-8-27b-iq4-xs-ryzen-ai-max-plus-395-128gb-lucebox-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
ryzen-ai-max-plus-395-128gb
id
qwen3-8-27b-iq4-xs-ryzen-ai-max-plus-395-128gb-lucebox-tp1
model instance id
unsloth-qwen3-8-27b-gguf--iq4-xs
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-8-27b-iq4-xs-ryzen-ai-max-plus-395-128gb-lucebox-tp1-sweep
status
candidate

capabilities

engine

name
lucebox
version
f686c447f067a04ea100a996e4c826e8cc4decc1

serving

max context tokens
131,072
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
Ryzen AI Max 395
localmaxxing · notes
Lucebox-native DFlash2 Qwen3.8-27B target. Task: count_to_30. One discarded warmup + three measured requests, median rates/timing; greedy decode (temperature omitted); target 27-token prompt -> 110 output tokens; 131072 context; batch 1. Target IQ4_XS GGUF with Q8_0 DFlash2 draft, DFlash2 spec-decode confirmed in server journal; identical output hashes across measured repetitions. Placement: Radeon 8060S iGPU (gfx1151) at PCIe 0000:c7:00.0, unified memory, weights streamed from system RAM; Q8 K/V cache. Peak VRAM is the maximum dflash_server DRM-client workload allocation (vram+gtt) on 0000:c7:00.0; power is peak observed GPU power. tokSOut uses server decode timing, tokSPrefill server prefill timing, tokSTotal client dispatch-to-DONE wall time.
localmaxxing · observed command
HIP_VISIBLE_DEVICES=1 /home/soulrider4ever/lucebox-src/server/build-hip-dual/dflash_server /home/soulrider4ever/lucebox-src/server/models/Qwen3.8-27B-UD-IQ4_XS.gguf --draft /home/soulrider4ever/lucebox-src/server/models/draft/qwen38-dflash2-q8_0.gguf --draft-block-size 16 --max-ctx 131072 --cache-type-k q8_0 --cache-type-v q8_0 --host 0.0.0.0 --port 8087 --target-device hip:0 --draft-device hip:0
localmaxxing · run id
cmt7on4h50059nn01bkp53rrp
localmaxxing · tokenized · environment · HIP VISIBLE DEVICES
1
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/cmt7on4h50059nn01bkp53rrp