Recipe

qwen3-5-9b-base-q4-k-xl-rtx-3060-12gb-llama-cpp-tp1

qwen3-5-9b-base-q4-k-xl-rtx-3060-12gb-llama-cpp-tp1

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
b10102 / 85e22ea0b
Accelerators
1
Tensor parallel
1
Context tokens
262,144
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/Qwen/Qwen3.5-9B
Repository
Qwen/Qwen3.5-9B
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/cmrjitcih02jvmj0173biwp25

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-bench
  2. -m
  3. /mnt/hdd/models/Qwen3.5-9B-UD-Q4_K_XL.gguf
  4. -p
  5. 512
  6. -n
  7. 128
  8. -b
  9. 4096
  10. -ub
  11. 1024
  12. -t
  13. 16
  14. -ngl
  15. 99
  16. -fa
  17. on
  18. -r
  19. 2
  20. -o
  21. json
FlagValue
-m/mnt/hdd/models/Qwen3.5-9B-UD-Q4_K_XL.gguf
-p512
-n128
-b4096
-ub1024
-t16
-ngl99
-faon
-r2
-ojson

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
1262,1441,588.147.6343.4observedqwen3-5-9b-base-q4-k-xl-rtx-3060-12gb-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
rtx-3060-12gb
id
qwen3-5-9b-base-q4-k-xl-rtx-3060-12gb-llama-cpp-tp1
model instance id
qwen-qwen3-5-9b--q4-k-xl
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-5-9b-base-q4-k-xl-rtx-3060-12gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
b10102 / 85e22ea0b

serving

max context tokens
262,144
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
cuda
localmaxxing · batch size
4,096
localmaxxing · hardware label
RTX 3060
localmaxxing · notes
Qwen3.5-9B UD Q4_K_XL, fully offloaded on RTX 3060. Maximum tested context 262144; server loaded and completed a real completion at that context. BeeLlama build 85e22ea0b build 10102, CUDA, Flash Attention, f16 benchmark KV cache. Optimization sweep tested batch/ubatch/thread combinations; winner b4096 ub1024 threads16. Peak VRAM measured after warmup at 262144 context with q4_0 KV cache.
localmaxxing · observed command
llama-bench -m /mnt/hdd/models/Qwen3.5-9B-UD-Q4_K_XL.gguf -p 512 -n 128 -b 4096 -ub 1024 -t 16 -ngl 99 -fa on -r 2 -o json
localmaxxing · run id
cmrjitcih02jvmj0173biwp25
localmaxxing · tokenized · arguments
llama-bench, -m, /mnt/hdd/models/Qwen3.5-9B-UD-Q4_K_XL.gguf, -p, 512, -n, 128, -b, 4096, -ub, 1024, -t, 16, -ngl, 99, -fa, on, -r, 2, -o, json
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/cmrjitcih02jvmj0173biwp25