Local AI Registry

Recipe

qwen3-6-35b-a3b-q4-k-xl-rx-9070-xt-16gb-llama-cpp-tp1

qwen3-6-35b-a3b-q4-k-xl-rx-9070-xt-16gb-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
b9174
Accelerators
1
Tensor parallel
1
Context tokens
131,072
Max concurrency
512
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/Qwen/Qwen3.6-35B-A3B
Repository
Qwen/Qwen3.6-35B-A3B
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/cmp7ymlnl00j1o401tduzlt7q

Observed source tokens

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

  1. sudo
  2. ./llama-server
  3. --threads
  4. 12
  5. --ctx-size
  6. 131072
  7. --flash-attn
  8. on
  9. --fit
  10. on
  11. --main-gpu
  12. 0
  13. --split-mode
  14. none
  15. --context-shift
  16. --cache-type-k
  17. q8_0
  18. --cache-type-v
  19. q8_0
  20. --n-cpu-moe
  21. 22
  22. --reasoning
  23. off
  24. --parallel
  25. 1
  26. --cache-ram
  27. 4096
  28. --temp
  29. 0.7
  30. --top-p
  31. 0.95
  32. --top-k
  33. 20
  34. --repeat-penalty
  35. 1.00
  36. --min-p
  37. 0.00
  38. --model
  39. /mnt/816d99c9-d5f0-40ae-96fd-b44f7f62faa4/misc/models/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf
  40. --mmproj
  41. /mnt/816d99c9-d5f0-40ae-96fd-b44f7f62faa4/misc/models/Qwen3.6-35B-A3B-mmproj-F16.gguf
  42. --host
  43. 192.168.1.102
  44. --port
  45. 8080
FlagValue
--threads12
--ctx-size131072
--flash-attnon
--fiton
--main-gpu0
--split-modenone
--cache-type-kq8_0
--cache-type-vq8_0
--n-cpu-moe22
--reasoningoff
--parallel1
--cache-ram4096
--temp0.7
--top-p0.95
--top-k20
--repeat-penalty1.00
--min-p0.00
--model/mnt/816d99c9-d5f0-40ae-96fd-b44f7f62faa4/misc/models/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf
--mmproj/mnt/816d99c9-d5f0-40ae-96fd-b44f7f62faa4/misc/models/Qwen3.6-35B-A3B-mmproj-F16.gguf
--host192.168.1.102
--port8080

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
512131,07239947317observedqwen3-6-35b-a3b-q4-k-xl-rx-9070-xt-16gb-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
rx-9070-xt-16gb
id
qwen3-6-35b-a3b-q4-k-xl-rx-9070-xt-16gb-llama-cpp-tp1
model instance id
qwen-qwen3-6-35b-a3b--q4-k-xl
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-6-35b-a3b-q4-k-xl-rx-9070-xt-16gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
b9174

serving

max concurrency
512
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

metadata

localmaxxing · backend
rocm
localmaxxing · hardware label
RX 9070
localmaxxing · notes
ROCm 7.2.2, test after some HIP fixes and speedups were merged. Starts at 49 tok/s near the start and drops down slowly to ~47. No longer seeing any hitching during generation. Also note that power consumption during idle is ~80w which can be fixed by setting GPU_MAX_HW_QUEUES=1 but it dropped tok/s to more like 37 :/
localmaxxing · observed command
sudo ./llama-server --threads 12 --ctx-size 131072 --flash-attn on --fit on --main-gpu 0 --split-mode none --context-shift --cache-type-k q8_0 --cache-type-v q8_0 --n-cpu-moe 22 --reasoning off --parallel 1 --cache-ram 4096 --temp 0.7 --top-p 0.95 --top-k 20 --repeat-penalty 1.00 --min-p 0.00 --model /mnt/816d99c9-d5f0-40ae-96fd-b44f7f62faa4/misc/models/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf --mmproj /mnt/816d99c9-d5f0-40ae-96fd-b44f7f62faa4/misc/models/Qwen3.6-35B-A3B-mmproj-F16.gguf --host 192.168.1.102 --port 8080
localmaxxing · run id
cmp7ymlnl00j1o401tduzlt7q
localmaxxing · tokenized · arguments
sudo, ./llama-server, --threads, 12, --ctx-size, 131072, --flash-attn, on, --fit, on, --main-gpu, 0, --split-mode, none, --context-shift, --cache-type-k, q8_0, --cache-type-v, q8_0, --n-cpu-moe, 22, --reasoning, off, --parallel, 1, --cache-ram, 4096, --temp, 0.7, --top-p, 0.95, --top-k, 20, --repeat-penalty, 1.00, --min-p, 0.00, --model, /mnt/816d99c9-d5f0-40ae-96fd-b44f7f62faa4/misc/models/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf, --mmproj, /mnt/816d99c9-d5f0-40ae-96fd-b44f7f62faa4/misc/models/Qwen3.6-35B-A3B-mmproj-F16.gguf, --host, 192.168.1.102, --port, 8080
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/cmp7ymlnl00j1o401tduzlt7q ↗