Local AI Registry

Recipe

qwen3-8-flash-next-iq4-xs-radeon-ai-pro-r9700-32gb-llama-cpp-tp1

qwen3-8-flash-next-iq4-xs-radeon-ai-pro-r9700-32gb-llama-cpp-tp1
JSON APIFind compatible recipes

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
035e227 + Qwen4Exp graph-reuse/QSA threshold patch 0249e47
Accelerators
1
Tensor parallel
1
Context tokens
32,768
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF
Repository
unsloth/Qwen3.8-Flash-Next-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/cmtbqjvpj003bqq01595qg0db

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. Qwen3.8-Flash-Next-UD-IQ4_XS-00001-of-00003.gguf
  4. --device
  5. ROCm0,ROCm1
  6. --tensor-split
  7. 0.45,0.55
  8. --split-mode
  9. layer
  10. --main-gpu
  11. 1
  12. --n-gpu-layers
  13. 99
  14. --ctx-size
  15. 32768
  16. --parallel
  17. 1
  18. --threads
  19. 32
  20. --threads-batch
  21. 32
  22. --ubatch-size
  23. 1024
  24. --jinja
  25. --flash-attn
  26. on
  27. --load-mode
  28. none
  29. --spec-type
  30. none
  31. --reasoning
  32. off
FlagValue
-mQwen3.8-Flash-Next-UD-IQ4_XS-00001-of-00003.gguf
--deviceROCm0,ROCm1
--tensor-split0.45,0.55
--split-modelayer
--main-gpu1
--n-gpu-layers99
--ctx-size32768
--parallel1
--threads32
--threads-batch32
--ubatch-size1024
--flash-attnon
--load-modenone
--spec-typenone
--reasoningoff

Environment

VariableValue
LLAMA_QWEN4EXP_QSA_MIN_KV32768

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
132,768537.425.64,747.2observedqwen3-8-flash-next-iq4-xs-radeon-ai-pro-r9700-32gb-llama-cpp-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
radeon-ai-pro-r9700-32gb
id
qwen3-8-flash-next-iq4-xs-radeon-ai-pro-r9700-32gb-llama-cpp-tp1
model instance id
unsloth-qwen3-8-flash-next-gguf--iq4-xs
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-8-flash-next-iq4-xs-radeon-ai-pro-r9700-32gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
035e227 + Qwen4Exp graph-reuse/QSA threshold patch 0249e47

serving

max concurrency
1
max context tokens
32,768
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-derived-from-source-evidence
serving.max concurrency · state known

metadata

localmaxxing · backend
rocm
localmaxxing · batch size
1
localmaxxing · hardware label
Radeon AI Pro R9700
localmaxxing · notes
Semantically validated Lucebox dual-GPU run. The LocalMaxxing hardware schema records the discrete R9700; the Radeon 8060S unified-memory iGPU also executed the 55% layer share and is disclosed here. Three-shard Qwen3.8-Flash-Next IQ4_XS GGUF, 2549-token synthetic prompt, natural 87-token completion, temperature 0, batch/concurrency 1, 32K context, one discarded warmup plus three measured streams. Decode and prefill use server-authoritative timings; total is (prompt+output)/client dispatch-to-DONE wall time; TTFT is dispatch to first content token. Peak VRAM is the maximum R9700 workload allocation from unique amdgpu DRM fdinfo clients at PCI 0000:c5:00.0, not device capacity or combined system memory. All three outputs shared SHA-256 93231b0b1ec900896c2237684dadd72a894e2c56cc865b8c10f7a5a87d996c4d. Greeting, multi-turn recall, streaming content, and terminal DONE semantic canaries passed.
localmaxxing · observed command
LLAMA_QWEN4EXP_QSA_MIN_KV=32768 llama-server -m Qwen3.8-Flash-Next-UD-IQ4_XS-00001-of-00003.gguf --device ROCm0,ROCm1 --tensor-split 0.45,0.55 --split-mode layer --main-gpu 1 --n-gpu-layers 99 --ctx-size 32768 --parallel 1 --threads 32 --threads-batch 32 --ubatch-size 1024 --jinja --flash-attn on --load-mode none --spec-type none --reasoning off
localmaxxing · run id
cmtbqjvpj003bqq01595qg0db
localmaxxing · tokenized · arguments
llama-server, -m, Qwen3.8-Flash-Next-UD-IQ4_XS-00001-of-00003.gguf, --device, ROCm0,ROCm1, --tensor-split, 0.45,0.55, --split-mode, layer, --main-gpu, 1, --n-gpu-layers, 99, --ctx-size, 32768, --parallel, 1, --threads, 32, --threads-batch, 32, --ubatch-size, 1024, --jinja, --flash-attn, on, --load-mode, none, --spec-type, none, --reasoning, off
localmaxxing · tokenized · environment · LLAMA QWEN4EXP QSA MIN KV
32768
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/cmtbqjvpj003bqq01595qg0db ↗